diff --git "a/track-a.jsonl" "b/track-a.jsonl" new file mode 100644--- /dev/null +++ "b/track-a.jsonl" @@ -0,0 +1,243 @@ +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-managed-devices-llm` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-managed-devices-llm\n name: Report managed devices llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-managed-devices-llm, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-managed-devices-llm\n name: Report managed devices llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a devices agent for OpenAdminOS: read managed devices (fields: id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType; permission DeviceManagementManagedDevices.Read.All), then have the LLM summarize what an admin should check. Slug: check-managed-devices-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-managed-devices-llm\n name: Check managed devices llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-managed-devices-llm-scheduled` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-managed-devices-llm-scheduled\n name: Report managed devices llm scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-managed-devices-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-managed-devices-llm-scheduled\n name: Report managed devices llm scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-managed-devices-count-llm` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType, breaks them down with a count transform, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-managed-devices-count-llm\n name: Review managed devices count llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: count_by_compliancestate\n format: transform\n label: Count managed devices by complianceState\n detail: Breaks managed devices down by complianceState.\n settings:\n kind: count-by-field\n source: '{{ load_managed_devices.output }}'\n field: complianceState\n buckets:\n - compliant\n - noncompliant\n - unknown\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n count_by_compliancestate.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n breakdown: '{{ count_by_compliancestate.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-managed-devices-count-llm, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-managed-devices-count-llm\n name: Check managed devices count llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: count_by_operatingsystem\n format: transform\n label: Count managed devices by operatingSystem\n detail: Breaks managed devices down by operatingSystem.\n settings:\n kind: count-by-field\n source: '{{ load_managed_devices.output }}'\n field: operatingSystem\n buckets:\n - Windows\n - macOS\n - iOS\n - Android\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n count_by_operatingsystem.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n breakdown: '{{ count_by_operatingsystem.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-managed-devices-count-llm-scheduled` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType, breaks them down with a count transform, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-managed-devices-count-llm-scheduled\n name: Audit managed devices count llm scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: count_by_compliancestate\n format: transform\n label: Count managed devices by complianceState\n detail: Breaks managed devices down by complianceState.\n settings:\n kind: count-by-field\n source: '{{ load_managed_devices.output }}'\n field: complianceState\n buckets:\n - compliant\n - noncompliant\n - unknown\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n count_by_compliancestate.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n breakdown: '{{ count_by_compliancestate.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-managed-devices-count-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-managed-devices-count-llm-scheduled\n name: Check managed devices count llm scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: count_by_operatingsystem\n format: transform\n label: Count managed devices by operatingSystem\n detail: Breaks managed devices down by operatingSystem.\n settings:\n kind: count-by-field\n source: '{{ load_managed_devices.output }}'\n field: operatingSystem\n buckets:\n - Windows\n - macOS\n - iOS\n - Android\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n count_by_operatingsystem.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n breakdown: '{{ count_by_operatingsystem.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-managed-devices-sort-llm` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType, samples the oldest by lastSyncDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-managed-devices-sort-llm\n name: Check managed devices sort llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample managed devices by lastSyncDateTime\n detail: Lists managed devices ordered by lastSyncDateTime.\n settings:\n kind: sort-by\n source: '{{ load_managed_devices.output }}'\n field: lastSyncDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-managed-devices-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-managed-devices-sort-llm\n name: Check managed devices sort llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample managed devices by lastSyncDateTime\n detail: Lists managed devices ordered by lastSyncDateTime.\n settings:\n kind: sort-by\n source: '{{ load_managed_devices.output }}'\n field: lastSyncDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-managed-devices-sort-llm-scheduled` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType, samples the oldest by lastSyncDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-managed-devices-sort-llm-scheduled\n name: Review managed devices sort llm scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample managed devices by lastSyncDateTime\n detail: Lists managed devices ordered by lastSyncDateTime.\n settings:\n kind: sort-by\n source: '{{ load_managed_devices.output }}'\n field: lastSyncDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-managed-devices-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-managed-devices-sort-llm-scheduled\n name: Report managed devices sort llm scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample managed devices by lastSyncDateTime\n detail: Lists managed devices ordered by lastSyncDateTime.\n settings:\n kind: sort-by\n source: '{{ load_managed_devices.output }}'\n field: lastSyncDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n load_managed_devices.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-managed-devices-count-sort-llm` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType, breaks them down with a count transform, samples the oldest by lastSyncDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-managed-devices-count-sort-llm\n name: Check managed devices count sort llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: count_by_compliancestate\n format: transform\n label: Count managed devices by complianceState\n detail: Breaks managed devices down by complianceState.\n settings:\n kind: count-by-field\n source: '{{ load_managed_devices.output }}'\n field: complianceState\n buckets:\n - compliant\n - noncompliant\n - unknown\n - id: oldest_items\n format: transform\n label: Sample managed devices by lastSyncDateTime\n detail: Lists managed devices ordered by lastSyncDateTime.\n settings:\n kind: sort-by\n source: '{{ load_managed_devices.output }}'\n field: lastSyncDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n count_by_compliancestate.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n breakdown: '{{ count_by_compliancestate.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-managed-devices-count-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-managed-devices-count-sort-llm\n name: Report managed devices count sort llm\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: count_by_compliancestate\n format: transform\n label: Count managed devices by complianceState\n detail: Breaks managed devices down by complianceState.\n settings:\n kind: count-by-field\n source: '{{ load_managed_devices.output }}'\n field: complianceState\n buckets:\n - compliant\n - noncompliant\n - unknown\n - id: oldest_items\n format: transform\n label: Sample managed devices by lastSyncDateTime\n detail: Lists managed devices ordered by lastSyncDateTime.\n settings:\n kind: sort-by\n source: '{{ load_managed_devices.output }}'\n field: lastSyncDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n count_by_compliancestate.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n breakdown: '{{ count_by_compliancestate.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-managed-devices-count-sort-llm-scheduled` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType, breaks them down with a count transform, samples the oldest by lastSyncDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-managed-devices-count-sort-llm-scheduled\n name: Report managed devices count sort llm scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: count_by_operatingsystem\n format: transform\n label: Count managed devices by operatingSystem\n detail: Breaks managed devices down by operatingSystem.\n settings:\n kind: count-by-field\n source: '{{ load_managed_devices.output }}'\n field: operatingSystem\n buckets:\n - Windows\n - macOS\n - iOS\n - Android\n - id: oldest_items\n format: transform\n label: Sample managed devices by lastSyncDateTime\n detail: Lists managed devices ordered by lastSyncDateTime.\n settings:\n kind: sort-by\n source: '{{ load_managed_devices.output }}'\n field: lastSyncDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n count_by_operatingsystem.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n breakdown: '{{ count_by_operatingsystem.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-managed-devices-count-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-managed-devices-count-sort-llm-scheduled\n name: Check managed devices count sort llm scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: count_by_compliancestate\n format: transform\n label: Count managed devices by complianceState\n detail: Breaks managed devices down by complianceState.\n settings:\n kind: count-by-field\n source: '{{ load_managed_devices.output }}'\n field: complianceState\n buckets:\n - compliant\n - noncompliant\n - unknown\n - id: oldest_items\n format: transform\n label: Sample managed devices by lastSyncDateTime\n detail: Lists managed devices ordered by lastSyncDateTime.\n settings:\n kind: sort-by\n source: '{{ load_managed_devices.output }}'\n field: lastSyncDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short managed devices report.\n settings:\n system: >-\n You review managed devices for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total managed devices: {{ load_managed_devices.output | size }}. Details: {{\n count_by_compliancestate.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_managed_devices.output | size }}'\n breakdown: '{{ count_by_compliancestate.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-managed-devices-list` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-managed-devices-list\n name: Check managed devices list\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_managed_devices.output | size }} managed devices.\n data:\n total: '{{ load_managed_devices.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-managed-devices-list, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-managed-devices-list\n name: Audit managed devices list\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_managed_devices.output | size }} managed devices.\n data:\n total: '{{ load_managed_devices.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-managed-devices-list-scheduled` that loads managed devices from Graph (/deviceManagement/managedDevices, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, deviceName, operatingSystem, complianceState, lastSyncDateTime, userPrincipalName, model, managedDeviceOwnerType. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-managed-devices-list-scheduled\n name: Audit managed devices list scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_managed_devices.output | size }} managed devices.\n data:\n total: '{{ load_managed_devices.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-managed-devices-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's managed devices via /deviceManagement/managedDevices. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-managed-devices-list-scheduled\n name: Review managed devices list scheduled\n description: Reviews managed devices and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_managed_devices\n format: graph\n label: Load managed devices\n detail: Reads managed devices from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/managedDevices\n select:\n - id\n - deviceName\n - operatingSystem\n - complianceState\n - lastSyncDateTime\n - userPrincipalName\n - model\n - managedDeviceOwnerType\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_managed_devices.output | size }} managed devices.\n data:\n total: '{{ load_managed_devices.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-users-llm` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-users-llm\n name: Report users llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-users-llm, version 0.1.0, author Community/community) that reviews the tenant's users via /users and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-users-llm\n name: Report users llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read users (fields: id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime; permission User.Read.All), then have the LLM summarize what an admin should check. Slug: report-users-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-users-llm\n name: Report users llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-users-llm-scheduled` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-users-llm-scheduled\n name: Check users llm scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-users-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's users via /users and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-users-llm-scheduled\n name: Report users llm scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-users-count-llm` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime, breaks them down with a count transform, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-users-count-llm\n name: Check users count llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: count_by_accountenabled\n format: transform\n label: Count users by accountEnabled\n detail: Breaks users down by accountEnabled.\n settings:\n kind: count-by-field\n source: '{{ load_users.output }}'\n field: accountEnabled\n buckets:\n - 'true'\n - 'false'\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ count_by_accountenabled.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n breakdown: '{{ count_by_accountenabled.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-users-count-llm, version 0.1.0, author Community/community) that reviews the tenant's users via /users with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-users-count-llm\n name: Review users count llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: count_by_accountenabled\n format: transform\n label: Count users by accountEnabled\n detail: Breaks users down by accountEnabled.\n settings:\n kind: count-by-field\n source: '{{ load_users.output }}'\n field: accountEnabled\n buckets:\n - 'true'\n - 'false'\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ count_by_accountenabled.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n breakdown: '{{ count_by_accountenabled.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-users-count-llm-scheduled` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime, breaks them down with a count transform, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-users-count-llm-scheduled\n name: Review users count llm scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: count_by_usertype\n format: transform\n label: Count users by userType\n detail: Breaks users down by userType.\n settings:\n kind: count-by-field\n source: '{{ load_users.output }}'\n field: userType\n buckets:\n - Member\n - Guest\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ count_by_usertype.output }}. Write\n a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n breakdown: '{{ count_by_usertype.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-users-count-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's users via /users with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-users-count-llm-scheduled\n name: Audit users count llm scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: count_by_accountenabled\n format: transform\n label: Count users by accountEnabled\n detail: Breaks users down by accountEnabled.\n settings:\n kind: count-by-field\n source: '{{ load_users.output }}'\n field: accountEnabled\n buckets:\n - 'true'\n - 'false'\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ count_by_accountenabled.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n breakdown: '{{ count_by_accountenabled.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-users-sort-llm` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime, samples the oldest by createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-users-sort-llm\n name: Report users sort llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-users-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's users via /users and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-users-sort-llm\n name: Audit users sort llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read users (fields: id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime; permission User.Read.All), show the oldest ten by createdDateTime, then have the LLM summarize what an admin should check. Slug: review-users-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-users-sort-llm\n name: Review users sort llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-users-sort-llm-scheduled` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime, samples the oldest by createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-users-sort-llm-scheduled\n name: Check users sort llm scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-users-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's users via /users and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-users-sort-llm-scheduled\n name: Check users sort llm scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ load_users.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-users-count-sort-llm` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime, breaks them down with a count transform, samples the oldest by createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-users-count-sort-llm\n name: Check users count sort llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: count_by_accountenabled\n format: transform\n label: Count users by accountEnabled\n detail: Breaks users down by accountEnabled.\n settings:\n kind: count-by-field\n source: '{{ load_users.output }}'\n field: accountEnabled\n buckets:\n - 'true'\n - 'false'\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ count_by_accountenabled.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n breakdown: '{{ count_by_accountenabled.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-users-count-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's users via /users with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-users-count-sort-llm\n name: Audit users count sort llm\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: count_by_usertype\n format: transform\n label: Count users by userType\n detail: Breaks users down by userType.\n settings:\n kind: count-by-field\n source: '{{ load_users.output }}'\n field: userType\n buckets:\n - Member\n - Guest\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ count_by_usertype.output }}. Write\n a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n breakdown: '{{ count_by_usertype.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-users-count-sort-llm-scheduled` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime, breaks them down with a count transform, samples the oldest by createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-users-count-sort-llm-scheduled\n name: Audit users count sort llm scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: count_by_accountenabled\n format: transform\n label: Count users by accountEnabled\n detail: Breaks users down by accountEnabled.\n settings:\n kind: count-by-field\n source: '{{ load_users.output }}'\n field: accountEnabled\n buckets:\n - 'true'\n - 'false'\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ count_by_accountenabled.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n breakdown: '{{ count_by_accountenabled.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-users-count-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's users via /users with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-users-count-sort-llm-scheduled\n name: Check users count sort llm scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\n - id: count_by_accountenabled\n format: transform\n label: Count users by accountEnabled\n detail: Breaks users down by accountEnabled.\n settings:\n kind: count-by-field\n source: '{{ load_users.output }}'\n field: accountEnabled\n buckets:\n - 'true'\n - 'false'\n - id: oldest_items\n format: transform\n label: Sample users by createdDateTime\n detail: Lists users ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_users.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short users report.\n settings:\n system: >-\n You review users for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total users: {{ load_users.output | size }}. Details: {{ count_by_accountenabled.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_users.output | size }}'\n breakdown: '{{ count_by_accountenabled.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-users-list` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-users-list\n name: Report users list\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_users.output | size }} users.\n data:\n total: '{{ load_users.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-users-list, version 0.1.0, author Community/community) that reviews the tenant's users via /users."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-users-list\n name: Review users list\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_users.output | size }} users.\n data:\n total: '{{ load_users.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read users (fields: id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime; permission User.Read.All). Slug: review-users-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-users-list\n name: Review users list\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_users.output | size }} users.\n data:\n total: '{{ load_users.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-users-list-scheduled` that loads users from Graph (/users, GET, scope User.Read.All) selecting id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-users-list-scheduled\n name: Check users list scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_users.output | size }} users.\n data:\n total: '{{ load_users.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-users-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's users via /users. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-users-list-scheduled\n name: Audit users list scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_users.output | size }} users.\n data:\n total: '{{ load_users.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read users (fields: id, displayName, userPrincipalName, accountEnabled, userType, createdDateTime; permission User.Read.All). Slug: report-users-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-users-list-scheduled\n name: Report users list scheduled\n description: Reviews users and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_users\n format: graph\n label: Load users\n detail: Reads users from the active tenant.\n settings:\n method: GET\n path: /users\n select:\n - id\n - displayName\n - userPrincipalName\n - accountEnabled\n - userType\n - createdDateTime\n scopes:\n - User.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_users.output | size }} users.\n data:\n total: '{{ load_users.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-groups-llm` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-groups-llm\n name: Audit groups llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-groups-llm, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-llm\n name: Review groups llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read groups (fields: id, displayName, groupTypes, membershipRule, createdDateTime; permission Group.Read.All), then have the LLM summarize what an admin should check. Slug: audit-groups-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-groups-llm\n name: Audit groups llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-groups-llm-scheduled` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-llm-scheduled\n name: Review groups llm scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-groups-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-groups-llm-scheduled\n name: Audit groups llm scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-groups-count-llm` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime, breaks them down with a count transform, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-count-llm\n name: Review groups count llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: count_by_grouptypes\n format: transform\n label: Count groups by groupTypes\n detail: Breaks groups down by groupTypes.\n settings:\n kind: count-by-field\n source: '{{ load_groups.output }}'\n field: groupTypes\n buckets:\n - Unified\n - DynamicMembership\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ count_by_grouptypes.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n breakdown: '{{ count_by_grouptypes.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-groups-count-llm, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-groups-count-llm\n name: Audit groups count llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: count_by_grouptypes\n format: transform\n label: Count groups by groupTypes\n detail: Breaks groups down by groupTypes.\n settings:\n kind: count-by-field\n source: '{{ load_groups.output }}'\n field: groupTypes\n buckets:\n - Unified\n - DynamicMembership\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ count_by_grouptypes.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n breakdown: '{{ count_by_grouptypes.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-groups-count-llm-scheduled` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime, breaks them down with a count transform, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-groups-count-llm-scheduled\n name: Check groups count llm scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: count_by_grouptypes\n format: transform\n label: Count groups by groupTypes\n detail: Breaks groups down by groupTypes.\n settings:\n kind: count-by-field\n source: '{{ load_groups.output }}'\n field: groupTypes\n buckets:\n - Unified\n - DynamicMembership\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ count_by_grouptypes.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n breakdown: '{{ count_by_grouptypes.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-groups-count-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-groups-count-llm-scheduled\n name: Report groups count llm scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: count_by_grouptypes\n format: transform\n label: Count groups by groupTypes\n detail: Breaks groups down by groupTypes.\n settings:\n kind: count-by-field\n source: '{{ load_groups.output }}'\n field: groupTypes\n buckets:\n - Unified\n - DynamicMembership\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ count_by_grouptypes.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n breakdown: '{{ count_by_grouptypes.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-groups-sort-llm` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime, samples the oldest by createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-groups-sort-llm\n name: Audit groups sort llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-groups-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-groups-sort-llm\n name: Check groups sort llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read groups (fields: id, displayName, groupTypes, membershipRule, createdDateTime; permission Group.Read.All), show the oldest ten by createdDateTime, then have the LLM summarize what an admin should check. Slug: report-groups-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-groups-sort-llm\n name: Report groups sort llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-groups-sort-llm-scheduled` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime, samples the oldest by createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-sort-llm-scheduled\n name: Review groups sort llm scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-groups-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-groups-sort-llm-scheduled\n name: Audit groups sort llm scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ load_groups.output }}. Write a\n compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-groups-count-sort-llm` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime, breaks them down with a count transform, samples the oldest by createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-count-sort-llm\n name: Review groups count sort llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: count_by_grouptypes\n format: transform\n label: Count groups by groupTypes\n detail: Breaks groups down by groupTypes.\n settings:\n kind: count-by-field\n source: '{{ load_groups.output }}'\n field: groupTypes\n buckets:\n - Unified\n - DynamicMembership\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ count_by_grouptypes.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n breakdown: '{{ count_by_grouptypes.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-groups-count-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-groups-count-sort-llm\n name: Check groups count sort llm\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: count_by_grouptypes\n format: transform\n label: Count groups by groupTypes\n detail: Breaks groups down by groupTypes.\n settings:\n kind: count-by-field\n source: '{{ load_groups.output }}'\n field: groupTypes\n buckets:\n - Unified\n - DynamicMembership\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ count_by_grouptypes.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n breakdown: '{{ count_by_grouptypes.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-groups-count-sort-llm-scheduled` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime, breaks them down with a count transform, samples the oldest by createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-groups-count-sort-llm-scheduled\n name: Check groups count sort llm scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: count_by_grouptypes\n format: transform\n label: Count groups by groupTypes\n detail: Breaks groups down by groupTypes.\n settings:\n kind: count-by-field\n source: '{{ load_groups.output }}'\n field: groupTypes\n buckets:\n - Unified\n - DynamicMembership\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ count_by_grouptypes.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n breakdown: '{{ count_by_grouptypes.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-groups-count-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-count-sort-llm-scheduled\n name: Review groups count sort llm scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\n - id: count_by_grouptypes\n format: transform\n label: Count groups by groupTypes\n detail: Breaks groups down by groupTypes.\n settings:\n kind: count-by-field\n source: '{{ load_groups.output }}'\n field: groupTypes\n buckets:\n - Unified\n - DynamicMembership\n - id: oldest_items\n format: transform\n label: Sample groups by createdDateTime\n detail: Lists groups ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_groups.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short groups report.\n settings:\n system: >-\n You review groups for Microsoft 365 admins. Be concise, factual, and conservative. Use only\n the supplied data.\n prompt: >-\n Total groups: {{ load_groups.output | size }}. Details: {{ count_by_grouptypes.output }}.\n Write a compact report: main finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_groups.output | size }}'\n breakdown: '{{ count_by_grouptypes.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-groups-list` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-list\n name: Review groups list\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_groups.output | size }} groups.\n data:\n total: '{{ load_groups.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-groups-list, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-list\n name: Review groups list\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_groups.output | size }} groups.\n data:\n total: '{{ load_groups.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read groups (fields: id, displayName, groupTypes, membershipRule, createdDateTime; permission Group.Read.All). Slug: audit-groups-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-groups-list\n name: Audit groups list\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_groups.output | size }} groups.\n data:\n total: '{{ load_groups.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-groups-list-scheduled` that loads groups from Graph (/groups, GET, scope Group.Read.All) selecting id, displayName, groupTypes, membershipRule, createdDateTime. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-list-scheduled\n name: Review groups list scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_groups.output | size }} groups.\n data:\n total: '{{ load_groups.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-groups-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's groups via /groups. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-list-scheduled\n name: Review groups list scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_groups.output | size }} groups.\n data:\n total: '{{ load_groups.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read groups (fields: id, displayName, groupTypes, membershipRule, createdDateTime; permission Group.Read.All). Slug: review-groups-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-groups-list-scheduled\n name: Review groups list scheduled\n description: Reviews groups and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_groups\n format: graph\n label: Load groups\n detail: Reads groups from the active tenant.\n settings:\n method: GET\n path: /groups\n select:\n - id\n - displayName\n - groupTypes\n - membershipRule\n - createdDateTime\n scopes:\n - Group.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_groups.output | size }} groups.\n data:\n total: '{{ load_groups.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-application-registrations-llm` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-application-registrations-llm\n name: Review application registrations llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-application-registrations-llm, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-application-registrations-llm\n name: Check application registrations llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read application registrations (fields: id, displayName, appId, createdDateTime, signInAudience; permission Application.Read.All), then have the LLM summarize what an admin should check. Slug: audit-application-registrations-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-application-registrations-llm\n name: Audit application registrations llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-application-registrations-llm-scheduled` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-application-registrations-llm-scheduled\n name: Audit application registrations llm scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-application-registrations-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-application-registrations-llm-scheduled\n name: Audit application registrations llm scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-application-registrations-count-llm` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience, breaks them down with a count transform, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-application-registrations-count-llm\n name: Audit application registrations count llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: count_by_signinaudience\n format: transform\n label: Count application registrations by signInAudience\n detail: Breaks application registrations down by signInAudience.\n settings:\n kind: count-by-field\n source: '{{ load_application_registrations.output }}'\n field: signInAudience\n buckets:\n - AzureADMyOrg\n - AzureADMultipleOrgs\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ count_by_signinaudience.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n breakdown: '{{ count_by_signinaudience.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-application-registrations-count-llm, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-application-registrations-count-llm\n name: Report application registrations count llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: count_by_signinaudience\n format: transform\n label: Count application registrations by signInAudience\n detail: Breaks application registrations down by signInAudience.\n settings:\n kind: count-by-field\n source: '{{ load_application_registrations.output }}'\n field: signInAudience\n buckets:\n - AzureADMyOrg\n - AzureADMultipleOrgs\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ count_by_signinaudience.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n breakdown: '{{ count_by_signinaudience.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-application-registrations-count-llm-scheduled` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience, breaks them down with a count transform, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-application-registrations-count-llm-scheduled\n name: Review application registrations count llm scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: count_by_signinaudience\n format: transform\n label: Count application registrations by signInAudience\n detail: Breaks application registrations down by signInAudience.\n settings:\n kind: count-by-field\n source: '{{ load_application_registrations.output }}'\n field: signInAudience\n buckets:\n - AzureADMyOrg\n - AzureADMultipleOrgs\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ count_by_signinaudience.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n breakdown: '{{ count_by_signinaudience.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-application-registrations-count-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-application-registrations-count-llm-scheduled\n name: Report application registrations count llm scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: count_by_signinaudience\n format: transform\n label: Count application registrations by signInAudience\n detail: Breaks application registrations down by signInAudience.\n settings:\n kind: count-by-field\n source: '{{ load_application_registrations.output }}'\n field: signInAudience\n buckets:\n - AzureADMyOrg\n - AzureADMultipleOrgs\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ count_by_signinaudience.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n breakdown: '{{ count_by_signinaudience.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-application-registrations-sort-llm` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience, samples the oldest by createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-application-registrations-sort-llm\n name: Review application registrations sort llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-application-registrations-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-application-registrations-sort-llm\n name: Audit application registrations sort llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read application registrations (fields: id, displayName, appId, createdDateTime, signInAudience; permission Application.Read.All), show the oldest ten by createdDateTime, then have the LLM summarize what an admin should check. Slug: audit-application-registrations-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-application-registrations-sort-llm\n name: Audit application registrations sort llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-application-registrations-sort-llm-scheduled` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience, samples the oldest by createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-application-registrations-sort-llm-scheduled\n name: Review application registrations sort llm scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-application-registrations-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-application-registrations-sort-llm-scheduled\n name: Report application registrations sort llm scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ load_application_registrations.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-application-registrations-count-sort-llm` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience, breaks them down with a count transform, samples the oldest by createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-application-registrations-count-sort-llm\n name: Report application registrations count sort llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: count_by_signinaudience\n format: transform\n label: Count application registrations by signInAudience\n detail: Breaks application registrations down by signInAudience.\n settings:\n kind: count-by-field\n source: '{{ load_application_registrations.output }}'\n field: signInAudience\n buckets:\n - AzureADMyOrg\n - AzureADMultipleOrgs\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ count_by_signinaudience.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n breakdown: '{{ count_by_signinaudience.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-application-registrations-count-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-application-registrations-count-sort-llm\n name: Check application registrations count sort llm\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: count_by_signinaudience\n format: transform\n label: Count application registrations by signInAudience\n detail: Breaks application registrations down by signInAudience.\n settings:\n kind: count-by-field\n source: '{{ load_application_registrations.output }}'\n field: signInAudience\n buckets:\n - AzureADMyOrg\n - AzureADMultipleOrgs\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ count_by_signinaudience.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n breakdown: '{{ count_by_signinaudience.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-application-registrations-count-sort-llm-scheduled` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience, breaks them down with a count transform, samples the oldest by createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-application-registrations-count-sort-llm-scheduled\n name: Review application registrations count sort llm scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: count_by_signinaudience\n format: transform\n label: Count application registrations by signInAudience\n detail: Breaks application registrations down by signInAudience.\n settings:\n kind: count-by-field\n source: '{{ load_application_registrations.output }}'\n field: signInAudience\n buckets:\n - AzureADMyOrg\n - AzureADMultipleOrgs\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ count_by_signinaudience.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n breakdown: '{{ count_by_signinaudience.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-application-registrations-count-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-application-registrations-count-sort-llm-scheduled\n name: Report application registrations count sort llm scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\n - id: count_by_signinaudience\n format: transform\n label: Count application registrations by signInAudience\n detail: Breaks application registrations down by signInAudience.\n settings:\n kind: count-by-field\n source: '{{ load_application_registrations.output }}'\n field: signInAudience\n buckets:\n - AzureADMyOrg\n - AzureADMultipleOrgs\n - id: oldest_items\n format: transform\n label: Sample application registrations by createdDateTime\n detail: Lists application registrations ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_application_registrations.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short application registrations report.\n settings:\n system: >-\n You review application registrations for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total application registrations: {{ load_application_registrations.output | size }}.\n Details: {{ count_by_signinaudience.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_application_registrations.output | size }}'\n breakdown: '{{ count_by_signinaudience.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-application-registrations-list` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-application-registrations-list\n name: Review application registrations list\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_application_registrations.output | size }} application registrations.\n data:\n total: '{{ load_application_registrations.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-application-registrations-list, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-application-registrations-list\n name: Review application registrations list\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_application_registrations.output | size }} application registrations.\n data:\n total: '{{ load_application_registrations.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read application registrations (fields: id, displayName, appId, createdDateTime, signInAudience; permission Application.Read.All). Slug: check-application-registrations-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-application-registrations-list\n name: Check application registrations list\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_application_registrations.output | size }} application registrations.\n data:\n total: '{{ load_application_registrations.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-application-registrations-list-scheduled` that loads application registrations from Graph (/applications, GET, scope Application.Read.All) selecting id, displayName, appId, createdDateTime, signInAudience. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-application-registrations-list-scheduled\n name: Report application registrations list scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_application_registrations.output | size }} application registrations.\n data:\n total: '{{ load_application_registrations.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-application-registrations-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's application registrations via /applications. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-application-registrations-list-scheduled\n name: Audit application registrations list scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_application_registrations.output | size }} application registrations.\n data:\n total: '{{ load_application_registrations.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read application registrations (fields: id, displayName, appId, createdDateTime, signInAudience; permission Application.Read.All). Slug: audit-application-registrations-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-application-registrations-list-scheduled\n name: Audit application registrations list scheduled\n description: Reviews application registrations and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_application_registrations\n format: graph\n label: Load application registrations\n detail: Reads application registrations from the active tenant.\n settings:\n method: GET\n path: /applications\n select:\n - id\n - displayName\n - appId\n - createdDateTime\n - signInAudience\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_application_registrations.output | size }} application registrations.\n data:\n total: '{{ load_application_registrations.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-conditional-access-policies-llm` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-conditional-access-policies-llm\n name: Check conditional access policies llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-conditional-access-policies-llm, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-conditional-access-policies-llm\n name: Report conditional access policies llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read conditional access policies (fields: id, displayName, state, createdDateTime, modifiedDateTime; permission Policy.Read.All), then have the LLM summarize what an admin should check. Slug: audit-conditional-access-policies-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-conditional-access-policies-llm\n name: Audit conditional access policies llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-conditional-access-policies-llm-scheduled` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-conditional-access-policies-llm-scheduled\n name: Review conditional access policies llm scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-conditional-access-policies-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-conditional-access-policies-llm-scheduled\n name: Review conditional access policies llm scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-conditional-access-policies-count-llm` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime, breaks them down with a count transform, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-conditional-access-policies-count-llm\n name: Report conditional access policies count llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: count_by_state\n format: transform\n label: Count conditional access policies by state\n detail: Breaks conditional access policies down by state.\n settings:\n kind: count-by-field\n source: '{{ load_conditional_access_policies.output }}'\n field: state\n buckets:\n - enabled\n - disabled\n - enabledForReportingButNotEnforced\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ count_by_state.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n breakdown: '{{ count_by_state.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-conditional-access-policies-count-llm, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-conditional-access-policies-count-llm\n name: Audit conditional access policies count llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: count_by_state\n format: transform\n label: Count conditional access policies by state\n detail: Breaks conditional access policies down by state.\n settings:\n kind: count-by-field\n source: '{{ load_conditional_access_policies.output }}'\n field: state\n buckets:\n - enabled\n - disabled\n - enabledForReportingButNotEnforced\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ count_by_state.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n breakdown: '{{ count_by_state.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-conditional-access-policies-count-llm-scheduled` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime, breaks them down with a count transform, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-conditional-access-policies-count-llm-scheduled\n name: Report conditional access policies count llm scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: count_by_state\n format: transform\n label: Count conditional access policies by state\n detail: Breaks conditional access policies down by state.\n settings:\n kind: count-by-field\n source: '{{ load_conditional_access_policies.output }}'\n field: state\n buckets:\n - enabled\n - disabled\n - enabledForReportingButNotEnforced\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ count_by_state.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n breakdown: '{{ count_by_state.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-conditional-access-policies-count-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-conditional-access-policies-count-llm-scheduled\n name: Check conditional access policies count llm scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: count_by_state\n format: transform\n label: Count conditional access policies by state\n detail: Breaks conditional access policies down by state.\n settings:\n kind: count-by-field\n source: '{{ load_conditional_access_policies.output }}'\n field: state\n buckets:\n - enabled\n - disabled\n - enabledForReportingButNotEnforced\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ count_by_state.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n breakdown: '{{ count_by_state.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-conditional-access-policies-sort-llm` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime, samples the oldest by modifiedDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-conditional-access-policies-sort-llm\n name: Check conditional access policies sort llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-conditional-access-policies-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-conditional-access-policies-sort-llm\n name: Report conditional access policies sort llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read conditional access policies (fields: id, displayName, state, createdDateTime, modifiedDateTime; permission Policy.Read.All), show the oldest ten by modifiedDateTime, then have the LLM summarize what an admin should check. Slug: report-conditional-access-policies-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-conditional-access-policies-sort-llm\n name: Report conditional access policies sort llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-conditional-access-policies-sort-llm-scheduled` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime, samples the oldest by modifiedDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-conditional-access-policies-sort-llm-scheduled\n name: Audit conditional access policies sort llm scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-conditional-access-policies-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-conditional-access-policies-sort-llm-scheduled\n name: Report conditional access policies sort llm scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ load_conditional_access_policies.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-conditional-access-policies-count-sort-llm` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime, breaks them down with a count transform, samples the oldest by modifiedDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-conditional-access-policies-count-sort-llm\n name: Report conditional access policies count sort llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: count_by_state\n format: transform\n label: Count conditional access policies by state\n detail: Breaks conditional access policies down by state.\n settings:\n kind: count-by-field\n source: '{{ load_conditional_access_policies.output }}'\n field: state\n buckets:\n - enabled\n - disabled\n - enabledForReportingButNotEnforced\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ count_by_state.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n breakdown: '{{ count_by_state.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-conditional-access-policies-count-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-conditional-access-policies-count-sort-llm\n name: Audit conditional access policies count sort llm\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: count_by_state\n format: transform\n label: Count conditional access policies by state\n detail: Breaks conditional access policies down by state.\n settings:\n kind: count-by-field\n source: '{{ load_conditional_access_policies.output }}'\n field: state\n buckets:\n - enabled\n - disabled\n - enabledForReportingButNotEnforced\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ count_by_state.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n breakdown: '{{ count_by_state.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-conditional-access-policies-count-sort-llm-scheduled` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime, breaks them down with a count transform, samples the oldest by modifiedDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-conditional-access-policies-count-sort-llm-scheduled\n name: Audit conditional access policies count sort llm scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: count_by_state\n format: transform\n label: Count conditional access policies by state\n detail: Breaks conditional access policies down by state.\n settings:\n kind: count-by-field\n source: '{{ load_conditional_access_policies.output }}'\n field: state\n buckets:\n - enabled\n - disabled\n - enabledForReportingButNotEnforced\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ count_by_state.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n breakdown: '{{ count_by_state.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-conditional-access-policies-count-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-conditional-access-policies-count-sort-llm-scheduled\n name: Review conditional access policies count sort llm scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\n - id: count_by_state\n format: transform\n label: Count conditional access policies by state\n detail: Breaks conditional access policies down by state.\n settings:\n kind: count-by-field\n source: '{{ load_conditional_access_policies.output }}'\n field: state\n buckets:\n - enabled\n - disabled\n - enabledForReportingButNotEnforced\n - id: oldest_items\n format: transform\n label: Sample conditional access policies by modifiedDateTime\n detail: Lists conditional access policies ordered by modifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_conditional_access_policies.output }}'\n field: modifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short conditional access policies report.\n settings:\n system: >-\n You review conditional access policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total conditional access policies: {{ load_conditional_access_policies.output | size }}.\n Details: {{ count_by_state.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n breakdown: '{{ count_by_state.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-conditional-access-policies-list` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-conditional-access-policies-list\n name: Check conditional access policies list\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_conditional_access_policies.output | size }} conditional access policies.\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-conditional-access-policies-list, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-conditional-access-policies-list\n name: Review conditional access policies list\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_conditional_access_policies.output | size }} conditional access policies.\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read conditional access policies (fields: id, displayName, state, createdDateTime, modifiedDateTime; permission Policy.Read.All). Slug: check-conditional-access-policies-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-conditional-access-policies-list\n name: Check conditional access policies list\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_conditional_access_policies.output | size }} conditional access policies.\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-conditional-access-policies-list-scheduled` that loads conditional access policies from Graph (/identity/conditionalAccess/policies, GET, scope Policy.Read.All) selecting id, displayName, state, createdDateTime, modifiedDateTime. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-conditional-access-policies-list-scheduled\n name: Check conditional access policies list scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_conditional_access_policies.output | size }} conditional access policies.\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-conditional-access-policies-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's conditional access policies via /identity/conditionalAccess/policies. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-conditional-access-policies-list-scheduled\n name: Report conditional access policies list scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_conditional_access_policies.output | size }} conditional access policies.\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read conditional access policies (fields: id, displayName, state, createdDateTime, modifiedDateTime; permission Policy.Read.All). Slug: review-conditional-access-policies-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-conditional-access-policies-list-scheduled\n name: Review conditional access policies list scheduled\n description: Reviews conditional access policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_conditional_access_policies\n format: graph\n label: Load conditional access policies\n detail: Reads conditional access policies from the active tenant.\n settings:\n method: GET\n path: /identity/conditionalAccess/policies\n select:\n - id\n - displayName\n - state\n - createdDateTime\n - modifiedDateTime\n scopes:\n - Policy.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_conditional_access_policies.output | size }} conditional access policies.\n data:\n total: '{{ load_conditional_access_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-device-compliance-policies-llm` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-llm\n name: Report device compliance policies llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-device-compliance-policies-llm, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-compliance-policies-llm\n name: Check device compliance policies llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a devices agent for OpenAdminOS: read device compliance policies (fields: id, displayName, createdDateTime, lastModifiedDateTime, version; permission DeviceManagementConfiguration.Read.All), then have the LLM summarize what an admin should check. Slug: audit-device-compliance-policies-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-device-compliance-policies-llm\n name: Audit device compliance policies llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-device-compliance-policies-llm-scheduled` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-llm-scheduled\n name: Report device compliance policies llm scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-device-compliance-policies-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-compliance-policies-llm-scheduled\n name: Check device compliance policies llm scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-device-compliance-policies-count-llm` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version, breaks them down with a count transform, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-count-llm\n name: Report device compliance policies count llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: count_by_version\n format: transform\n label: Count device compliance policies by version\n detail: Breaks device compliance policies down by version.\n settings:\n kind: count-by-field\n source: '{{ load_device_compliance_policies.output }}'\n field: version\n buckets:\n - '1'\n - '2'\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ count_by_version.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n breakdown: '{{ count_by_version.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-device-compliance-policies-count-llm, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-device-compliance-policies-count-llm\n name: Review device compliance policies count llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: count_by_version\n format: transform\n label: Count device compliance policies by version\n detail: Breaks device compliance policies down by version.\n settings:\n kind: count-by-field\n source: '{{ load_device_compliance_policies.output }}'\n field: version\n buckets:\n - '1'\n - '2'\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ count_by_version.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n breakdown: '{{ count_by_version.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-device-compliance-policies-count-llm-scheduled` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version, breaks them down with a count transform, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-compliance-policies-count-llm-scheduled\n name: Check device compliance policies count llm scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: count_by_version\n format: transform\n label: Count device compliance policies by version\n detail: Breaks device compliance policies down by version.\n settings:\n kind: count-by-field\n source: '{{ load_device_compliance_policies.output }}'\n field: version\n buckets:\n - '1'\n - '2'\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ count_by_version.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n breakdown: '{{ count_by_version.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-device-compliance-policies-count-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-count-llm-scheduled\n name: Report device compliance policies count llm scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: count_by_version\n format: transform\n label: Count device compliance policies by version\n detail: Breaks device compliance policies down by version.\n settings:\n kind: count-by-field\n source: '{{ load_device_compliance_policies.output }}'\n field: version\n buckets:\n - '1'\n - '2'\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ count_by_version.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n breakdown: '{{ count_by_version.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-device-compliance-policies-sort-llm` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version, samples the oldest by lastModifiedDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-sort-llm\n name: Report device compliance policies sort llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-device-compliance-policies-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-sort-llm\n name: Report device compliance policies sort llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a devices agent for OpenAdminOS: read device compliance policies (fields: id, displayName, createdDateTime, lastModifiedDateTime, version; permission DeviceManagementConfiguration.Read.All), show the oldest ten by lastModifiedDateTime, then have the LLM summarize what an admin should check. Slug: report-device-compliance-policies-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-sort-llm\n name: Report device compliance policies sort llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-device-compliance-policies-sort-llm-scheduled` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version, samples the oldest by lastModifiedDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-sort-llm-scheduled\n name: Report device compliance policies sort llm scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-device-compliance-policies-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-compliance-policies-sort-llm-scheduled\n name: Check device compliance policies sort llm scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ load_device_compliance_policies.output }}. Write a compact report: main finding,\n what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-device-compliance-policies-count-sort-llm` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version, breaks them down with a count transform, samples the oldest by lastModifiedDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-compliance-policies-count-sort-llm\n name: Check device compliance policies count sort llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: count_by_version\n format: transform\n label: Count device compliance policies by version\n detail: Breaks device compliance policies down by version.\n settings:\n kind: count-by-field\n source: '{{ load_device_compliance_policies.output }}'\n field: version\n buckets:\n - '1'\n - '2'\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ count_by_version.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n breakdown: '{{ count_by_version.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-device-compliance-policies-count-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-compliance-policies-count-sort-llm\n name: Check device compliance policies count sort llm\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: count_by_version\n format: transform\n label: Count device compliance policies by version\n detail: Breaks device compliance policies down by version.\n settings:\n kind: count-by-field\n source: '{{ load_device_compliance_policies.output }}'\n field: version\n buckets:\n - '1'\n - '2'\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ count_by_version.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n breakdown: '{{ count_by_version.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-device-compliance-policies-count-sort-llm-scheduled` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version, breaks them down with a count transform, samples the oldest by lastModifiedDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-device-compliance-policies-count-sort-llm-scheduled\n name: Audit device compliance policies count sort llm scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: count_by_version\n format: transform\n label: Count device compliance policies by version\n detail: Breaks device compliance policies down by version.\n settings:\n kind: count-by-field\n source: '{{ load_device_compliance_policies.output }}'\n field: version\n buckets:\n - '1'\n - '2'\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ count_by_version.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n breakdown: '{{ count_by_version.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-device-compliance-policies-count-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-device-compliance-policies-count-sort-llm-scheduled\n name: Audit device compliance policies count sort llm scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: count_by_version\n format: transform\n label: Count device compliance policies by version\n detail: Breaks device compliance policies down by version.\n settings:\n kind: count-by-field\n source: '{{ load_device_compliance_policies.output }}'\n field: version\n buckets:\n - '1'\n - '2'\n - id: oldest_items\n format: transform\n label: Sample device compliance policies by lastModifiedDateTime\n detail: Lists device compliance policies ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_compliance_policies.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device compliance policies report.\n settings:\n system: >-\n You review device compliance policies for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device compliance policies: {{ load_device_compliance_policies.output | size }}.\n Details: {{ count_by_version.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n breakdown: '{{ count_by_version.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-device-compliance-policies-list` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-device-compliance-policies-list\n name: Audit device compliance policies list\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_device_compliance_policies.output | size }} device compliance policies.\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-device-compliance-policies-list, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-device-compliance-policies-list\n name: Audit device compliance policies list\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_device_compliance_policies.output | size }} device compliance policies.\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a devices agent for OpenAdminOS: read device compliance policies (fields: id, displayName, createdDateTime, lastModifiedDateTime, version; permission DeviceManagementConfiguration.Read.All). Slug: report-device-compliance-policies-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-compliance-policies-list\n name: Report device compliance policies list\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_device_compliance_policies.output | size }} device compliance policies.\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-device-compliance-policies-list-scheduled` that loads device compliance policies from Graph (/deviceManagement/deviceCompliancePolicies, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, version. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-compliance-policies-list-scheduled\n name: Check device compliance policies list scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_device_compliance_policies.output | size }} device compliance policies.\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-device-compliance-policies-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's device compliance policies via /deviceManagement/deviceCompliancePolicies. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-device-compliance-policies-list-scheduled\n name: Review device compliance policies list scheduled\n description: Reviews device compliance policies and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_compliance_policies\n format: graph\n label: Load device compliance policies\n detail: Reads device compliance policies from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceCompliancePolicies\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n - version\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_device_compliance_policies.output | size }} device compliance policies.\n data:\n total: '{{ load_device_compliance_policies.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-device-configuration-profiles-llm` that loads device configuration profiles from Graph (/deviceManagement/deviceConfigurations, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-configuration-profiles-llm\n name: Check device configuration profiles llm\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-device-configuration-profiles-llm, version 0.1.0, author Community/community) that reviews the tenant's device configuration profiles via /deviceManagement/deviceConfigurations and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-device-configuration-profiles-llm\n name: Review device configuration profiles llm\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a devices agent for OpenAdminOS: read device configuration profiles (fields: id, displayName, createdDateTime, lastModifiedDateTime; permission DeviceManagementConfiguration.Read.All), then have the LLM summarize what an admin should check. Slug: audit-device-configuration-profiles-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-device-configuration-profiles-llm\n name: Audit device configuration profiles llm\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-device-configuration-profiles-llm-scheduled` that loads device configuration profiles from Graph (/deviceManagement/deviceConfigurations, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-configuration-profiles-llm-scheduled\n name: Check device configuration profiles llm scheduled\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-device-configuration-profiles-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's device configuration profiles via /deviceManagement/deviceConfigurations and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-device-configuration-profiles-llm-scheduled\n name: Audit device configuration profiles llm scheduled\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-device-configuration-profiles-sort-llm` that loads device configuration profiles from Graph (/deviceManagement/deviceConfigurations, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, samples the oldest by lastModifiedDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-configuration-profiles-sort-llm\n name: Report device configuration profiles sort llm\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device configuration profiles by lastModifiedDateTime\n detail: Lists device configuration profiles ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_configuration_profiles.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-device-configuration-profiles-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's device configuration profiles via /deviceManagement/deviceConfigurations and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-configuration-profiles-sort-llm\n name: Check device configuration profiles sort llm\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device configuration profiles by lastModifiedDateTime\n detail: Lists device configuration profiles ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_configuration_profiles.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a devices agent for OpenAdminOS: read device configuration profiles (fields: id, displayName, createdDateTime, lastModifiedDateTime; permission DeviceManagementConfiguration.Read.All), show the oldest ten by lastModifiedDateTime, then have the LLM summarize what an admin should check. Slug: review-device-configuration-profiles-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-device-configuration-profiles-sort-llm\n name: Review device configuration profiles sort llm\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device configuration profiles by lastModifiedDateTime\n detail: Lists device configuration profiles ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_configuration_profiles.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-device-configuration-profiles-sort-llm-scheduled` that loads device configuration profiles from Graph (/deviceManagement/deviceConfigurations, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime, samples the oldest by lastModifiedDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-configuration-profiles-sort-llm-scheduled\n name: Report device configuration profiles sort llm scheduled\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device configuration profiles by lastModifiedDateTime\n detail: Lists device configuration profiles ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_configuration_profiles.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-device-configuration-profiles-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's device configuration profiles via /deviceManagement/deviceConfigurations and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-configuration-profiles-sort-llm-scheduled\n name: Check device configuration profiles sort llm scheduled\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\n - id: oldest_items\n format: transform\n label: Sample device configuration profiles by lastModifiedDateTime\n detail: Lists device configuration profiles ordered by lastModifiedDateTime.\n settings:\n kind: sort-by\n source: '{{ load_device_configuration_profiles.output }}'\n field: lastModifiedDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short device configuration profiles report.\n settings:\n system: >-\n You review device configuration profiles for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total device configuration profiles: {{ load_device_configuration_profiles.output | size }}.\n Details: {{ load_device_configuration_profiles.output }}. Write a compact report: main\n finding, what to verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-device-configuration-profiles-list` that loads device configuration profiles from Graph (/deviceManagement/deviceConfigurations, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-device-configuration-profiles-list\n name: Review device configuration profiles list\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_device_configuration_profiles.output | size }} device configuration profiles.\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-device-configuration-profiles-list, version 0.1.0, author Community/community) that reviews the tenant's device configuration profiles via /deviceManagement/deviceConfigurations."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-device-configuration-profiles-list\n name: Review device configuration profiles list\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_device_configuration_profiles.output | size }} device configuration profiles.\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a devices agent for OpenAdminOS: read device configuration profiles (fields: id, displayName, createdDateTime, lastModifiedDateTime; permission DeviceManagementConfiguration.Read.All). Slug: check-device-configuration-profiles-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-configuration-profiles-list\n name: Check device configuration profiles list\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_device_configuration_profiles.output | size }} device configuration profiles.\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-device-configuration-profiles-list-scheduled` that loads device configuration profiles from Graph (/deviceManagement/deviceConfigurations, GET, scope DeviceManagementConfiguration.Read.All) selecting id, displayName, createdDateTime, lastModifiedDateTime. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-configuration-profiles-list-scheduled\n name: Check device configuration profiles list scheduled\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_device_configuration_profiles.output | size }} device configuration profiles.\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-device-configuration-profiles-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's device configuration profiles via /deviceManagement/deviceConfigurations. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-device-configuration-profiles-list-scheduled\n name: Check device configuration profiles list scheduled\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_device_configuration_profiles.output | size }} device configuration profiles.\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a devices agent for OpenAdminOS: read device configuration profiles (fields: id, displayName, createdDateTime, lastModifiedDateTime; permission DeviceManagementConfiguration.Read.All). Slug: report-device-configuration-profiles-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-device-configuration-profiles-list-scheduled\n name: Report device configuration profiles list scheduled\n description: Reviews device configuration profiles and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: devices\n mode: read\nskills:\n - id: load_device_configuration_profiles\n format: graph\n label: Load device configuration profiles\n detail: Reads device configuration profiles from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/deviceConfigurations\n select:\n - id\n - displayName\n - createdDateTime\n - lastModifiedDateTime\n scopes:\n - DeviceManagementConfiguration.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_device_configuration_profiles.output | size }} device configuration profiles.\n data:\n total: '{{ load_device_configuration_profiles.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-detected-apps-llm` that loads detected apps from Graph (/deviceManagement/detectedApps, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, displayName, version, deviceCount, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-detected-apps-llm\n name: Audit detected apps llm\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-detected-apps-llm, version 0.1.0, author Community/community) that reviews the tenant's detected apps via /deviceManagement/detectedApps and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-detected-apps-llm\n name: Audit detected apps llm\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a apps agent for OpenAdminOS: read detected apps (fields: id, displayName, version, deviceCount; permission DeviceManagementManagedDevices.Read.All), then have the LLM summarize what an admin should check. Slug: audit-detected-apps-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-detected-apps-llm\n name: Audit detected apps llm\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-detected-apps-llm-scheduled` that loads detected apps from Graph (/deviceManagement/detectedApps, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, displayName, version, deviceCount, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-detected-apps-llm-scheduled\n name: Review detected apps llm scheduled\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-detected-apps-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's detected apps via /deviceManagement/detectedApps and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-detected-apps-llm-scheduled\n name: Check detected apps llm scheduled\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-detected-apps-sort-llm` that loads detected apps from Graph (/deviceManagement/detectedApps, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, displayName, version, deviceCount, samples the oldest by deviceCount, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-detected-apps-sort-llm\n name: Review detected apps sort llm\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample detected apps by deviceCount\n detail: Lists detected apps ordered by deviceCount.\n settings:\n kind: sort-by\n source: '{{ load_detected_apps.output }}'\n field: deviceCount\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-detected-apps-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's detected apps via /deviceManagement/detectedApps and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-detected-apps-sort-llm\n name: Report detected apps sort llm\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample detected apps by deviceCount\n detail: Lists detected apps ordered by deviceCount.\n settings:\n kind: sort-by\n source: '{{ load_detected_apps.output }}'\n field: deviceCount\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a apps agent for OpenAdminOS: read detected apps (fields: id, displayName, version, deviceCount; permission DeviceManagementManagedDevices.Read.All), show the oldest ten by deviceCount, then have the LLM summarize what an admin should check. Slug: audit-detected-apps-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-detected-apps-sort-llm\n name: Audit detected apps sort llm\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample detected apps by deviceCount\n detail: Lists detected apps ordered by deviceCount.\n settings:\n kind: sort-by\n source: '{{ load_detected_apps.output }}'\n field: deviceCount\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-detected-apps-sort-llm-scheduled` that loads detected apps from Graph (/deviceManagement/detectedApps, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, displayName, version, deviceCount, samples the oldest by deviceCount, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-detected-apps-sort-llm-scheduled\n name: Report detected apps sort llm scheduled\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample detected apps by deviceCount\n detail: Lists detected apps ordered by deviceCount.\n settings:\n kind: sort-by\n source: '{{ load_detected_apps.output }}'\n field: deviceCount\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-detected-apps-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's detected apps via /deviceManagement/detectedApps and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-detected-apps-sort-llm-scheduled\n name: Check detected apps sort llm scheduled\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\n - id: oldest_items\n format: transform\n label: Sample detected apps by deviceCount\n detail: Lists detected apps ordered by deviceCount.\n settings:\n kind: sort-by\n source: '{{ load_detected_apps.output }}'\n field: deviceCount\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short detected apps report.\n settings:\n system: >-\n You review detected apps for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total detected apps: {{ load_detected_apps.output | size }}. Details: {{\n load_detected_apps.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_detected_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-detected-apps-list` that loads detected apps from Graph (/deviceManagement/detectedApps, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, displayName, version, deviceCount. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-detected-apps-list\n name: Audit detected apps list\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_detected_apps.output | size }} detected apps.\n data:\n total: '{{ load_detected_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-detected-apps-list, version 0.1.0, author Community/community) that reviews the tenant's detected apps via /deviceManagement/detectedApps."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-detected-apps-list\n name: Check detected apps list\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_detected_apps.output | size }} detected apps.\n data:\n total: '{{ load_detected_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a apps agent for OpenAdminOS: read detected apps (fields: id, displayName, version, deviceCount; permission DeviceManagementManagedDevices.Read.All). Slug: review-detected-apps-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-detected-apps-list\n name: Review detected apps list\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_detected_apps.output | size }} detected apps.\n data:\n total: '{{ load_detected_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-detected-apps-list-scheduled` that loads detected apps from Graph (/deviceManagement/detectedApps, GET, scope DeviceManagementManagedDevices.Read.All) selecting id, displayName, version, deviceCount. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-detected-apps-list-scheduled\n name: Review detected apps list scheduled\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_detected_apps.output | size }} detected apps.\n data:\n total: '{{ load_detected_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-detected-apps-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's detected apps via /deviceManagement/detectedApps. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-detected-apps-list-scheduled\n name: Audit detected apps list scheduled\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_detected_apps.output | size }} detected apps.\n data:\n total: '{{ load_detected_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a apps agent for OpenAdminOS: read detected apps (fields: id, displayName, version, deviceCount; permission DeviceManagementManagedDevices.Read.All). Slug: review-detected-apps-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-detected-apps-list-scheduled\n name: Review detected apps list scheduled\n description: Reviews detected apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_detected_apps\n format: graph\n label: Load detected apps\n detail: Reads detected apps from the active tenant.\n settings:\n method: GET\n path: /deviceManagement/detectedApps\n select:\n - id\n - displayName\n - version\n - deviceCount\n scopes:\n - DeviceManagementManagedDevices.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_detected_apps.output | size }} detected apps.\n data:\n total: '{{ load_detected_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-sign-in-events-llm` that loads sign-in events from Graph (/auditLogs/signIns, GET, scope AuditLog.Read.All) selecting id, userPrincipalName, appDisplayName, createdDateTime, status, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-sign-in-events-llm\n name: Audit sign in events llm\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-sign-in-events-llm, version 0.1.0, author Community/community) that reviews the tenant's sign-in events via /auditLogs/signIns and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-sign-in-events-llm\n name: Check sign in events llm\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a compliance agent for OpenAdminOS: read sign-in events (fields: id, userPrincipalName, appDisplayName, createdDateTime, status; permission AuditLog.Read.All), then have the LLM summarize what an admin should check. Slug: check-sign-in-events-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-sign-in-events-llm\n name: Check sign in events llm\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-sign-in-events-llm-scheduled` that loads sign-in events from Graph (/auditLogs/signIns, GET, scope AuditLog.Read.All) selecting id, userPrincipalName, appDisplayName, createdDateTime, status, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-sign-in-events-llm-scheduled\n name: Check sign in events llm scheduled\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-sign-in-events-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's sign-in events via /auditLogs/signIns and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-sign-in-events-llm-scheduled\n name: Review sign in events llm scheduled\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-sign-in-events-sort-llm` that loads sign-in events from Graph (/auditLogs/signIns, GET, scope AuditLog.Read.All) selecting id, userPrincipalName, appDisplayName, createdDateTime, status, samples the oldest by createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-sign-in-events-sort-llm\n name: Audit sign in events sort llm\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample sign-in events by createdDateTime\n detail: Lists sign-in events ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_sign_in_events.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-sign-in-events-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's sign-in events via /auditLogs/signIns and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-sign-in-events-sort-llm\n name: Check sign in events sort llm\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample sign-in events by createdDateTime\n detail: Lists sign-in events ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_sign_in_events.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a compliance agent for OpenAdminOS: read sign-in events (fields: id, userPrincipalName, appDisplayName, createdDateTime, status; permission AuditLog.Read.All), show the oldest ten by createdDateTime, then have the LLM summarize what an admin should check. Slug: audit-sign-in-events-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-sign-in-events-sort-llm\n name: Audit sign in events sort llm\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample sign-in events by createdDateTime\n detail: Lists sign-in events ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_sign_in_events.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-sign-in-events-sort-llm-scheduled` that loads sign-in events from Graph (/auditLogs/signIns, GET, scope AuditLog.Read.All) selecting id, userPrincipalName, appDisplayName, createdDateTime, status, samples the oldest by createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-sign-in-events-sort-llm-scheduled\n name: Audit sign in events sort llm scheduled\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample sign-in events by createdDateTime\n detail: Lists sign-in events ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_sign_in_events.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-sign-in-events-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's sign-in events via /auditLogs/signIns and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-sign-in-events-sort-llm-scheduled\n name: Review sign in events sort llm scheduled\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample sign-in events by createdDateTime\n detail: Lists sign-in events ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_sign_in_events.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short sign-in events report.\n settings:\n system: >-\n You review sign-in events for Microsoft 365 admins. Be concise, factual, and conservative.\n Use only the supplied data.\n prompt: >-\n Total sign-in events: {{ load_sign_in_events.output | size }}. Details: {{\n load_sign_in_events.output }}. Write a compact report: main finding, what to verify, next\n actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_sign_in_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-sign-in-events-list` that loads sign-in events from Graph (/auditLogs/signIns, GET, scope AuditLog.Read.All) selecting id, userPrincipalName, appDisplayName, createdDateTime, status. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-sign-in-events-list\n name: Report sign in events list\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_sign_in_events.output | size }} sign-in events.\n data:\n total: '{{ load_sign_in_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-sign-in-events-list, version 0.1.0, author Community/community) that reviews the tenant's sign-in events via /auditLogs/signIns."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-sign-in-events-list\n name: Check sign in events list\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_sign_in_events.output | size }} sign-in events.\n data:\n total: '{{ load_sign_in_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a compliance agent for OpenAdminOS: read sign-in events (fields: id, userPrincipalName, appDisplayName, createdDateTime, status; permission AuditLog.Read.All). Slug: check-sign-in-events-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-sign-in-events-list\n name: Check sign in events list\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_sign_in_events.output | size }} sign-in events.\n data:\n total: '{{ load_sign_in_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-sign-in-events-list-scheduled` that loads sign-in events from Graph (/auditLogs/signIns, GET, scope AuditLog.Read.All) selecting id, userPrincipalName, appDisplayName, createdDateTime, status. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-sign-in-events-list-scheduled\n name: Review sign in events list scheduled\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_sign_in_events.output | size }} sign-in events.\n data:\n total: '{{ load_sign_in_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-sign-in-events-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's sign-in events via /auditLogs/signIns. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-sign-in-events-list-scheduled\n name: Report sign in events list scheduled\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_sign_in_events.output | size }} sign-in events.\n data:\n total: '{{ load_sign_in_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a compliance agent for OpenAdminOS: read sign-in events (fields: id, userPrincipalName, appDisplayName, createdDateTime, status; permission AuditLog.Read.All). Slug: audit-sign-in-events-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-sign-in-events-list-scheduled\n name: Audit sign in events list scheduled\n description: Reviews sign-in events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_sign_in_events\n format: graph\n label: Load sign-in events\n detail: Reads sign-in events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/signIns\n select:\n - id\n - userPrincipalName\n - appDisplayName\n - createdDateTime\n - status\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_sign_in_events.output | size }} sign-in events.\n data:\n total: '{{ load_sign_in_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-directory-audit-events-llm` that loads directory audit events from Graph (/auditLogs/directoryAudits, GET, scope AuditLog.Read.All) selecting id, activityDisplayName, initiatedBy, activityDateTime, result, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-directory-audit-events-llm\n name: Check directory audit events llm\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-directory-audit-events-llm, version 0.1.0, author Community/community) that reviews the tenant's directory audit events via /auditLogs/directoryAudits and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-directory-audit-events-llm\n name: Check directory audit events llm\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a compliance agent for OpenAdminOS: read directory audit events (fields: id, activityDisplayName, initiatedBy, activityDateTime, result; permission AuditLog.Read.All), then have the LLM summarize what an admin should check. Slug: check-directory-audit-events-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-directory-audit-events-llm\n name: Check directory audit events llm\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-directory-audit-events-llm-scheduled` that loads directory audit events from Graph (/auditLogs/directoryAudits, GET, scope AuditLog.Read.All) selecting id, activityDisplayName, initiatedBy, activityDateTime, result, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-directory-audit-events-llm-scheduled\n name: Report directory audit events llm scheduled\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-directory-audit-events-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's directory audit events via /auditLogs/directoryAudits and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-directory-audit-events-llm-scheduled\n name: Audit directory audit events llm scheduled\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-directory-audit-events-sort-llm` that loads directory audit events from Graph (/auditLogs/directoryAudits, GET, scope AuditLog.Read.All) selecting id, activityDisplayName, initiatedBy, activityDateTime, result, samples the oldest by activityDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-directory-audit-events-sort-llm\n name: Check directory audit events sort llm\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample directory audit events by activityDateTime\n detail: Lists directory audit events ordered by activityDateTime.\n settings:\n kind: sort-by\n source: '{{ load_directory_audit_events.output }}'\n field: activityDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-directory-audit-events-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's directory audit events via /auditLogs/directoryAudits and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-directory-audit-events-sort-llm\n name: Report directory audit events sort llm\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample directory audit events by activityDateTime\n detail: Lists directory audit events ordered by activityDateTime.\n settings:\n kind: sort-by\n source: '{{ load_directory_audit_events.output }}'\n field: activityDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a compliance agent for OpenAdminOS: read directory audit events (fields: id, activityDisplayName, initiatedBy, activityDateTime, result; permission AuditLog.Read.All), show the oldest ten by activityDateTime, then have the LLM summarize what an admin should check. Slug: audit-directory-audit-events-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-directory-audit-events-sort-llm\n name: Audit directory audit events sort llm\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample directory audit events by activityDateTime\n detail: Lists directory audit events ordered by activityDateTime.\n settings:\n kind: sort-by\n source: '{{ load_directory_audit_events.output }}'\n field: activityDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-directory-audit-events-sort-llm-scheduled` that loads directory audit events from Graph (/auditLogs/directoryAudits, GET, scope AuditLog.Read.All) selecting id, activityDisplayName, initiatedBy, activityDateTime, result, samples the oldest by activityDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-directory-audit-events-sort-llm-scheduled\n name: Review directory audit events sort llm scheduled\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample directory audit events by activityDateTime\n detail: Lists directory audit events ordered by activityDateTime.\n settings:\n kind: sort-by\n source: '{{ load_directory_audit_events.output }}'\n field: activityDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-directory-audit-events-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's directory audit events via /auditLogs/directoryAudits and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-directory-audit-events-sort-llm-scheduled\n name: Review directory audit events sort llm scheduled\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\n - id: oldest_items\n format: transform\n label: Sample directory audit events by activityDateTime\n detail: Lists directory audit events ordered by activityDateTime.\n settings:\n kind: sort-by\n source: '{{ load_directory_audit_events.output }}'\n field: activityDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short directory audit events report.\n settings:\n system: >-\n You review directory audit events for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total directory audit events: {{ load_directory_audit_events.output | size }}. Details: {{\n load_directory_audit_events.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-directory-audit-events-list` that loads directory audit events from Graph (/auditLogs/directoryAudits, GET, scope AuditLog.Read.All) selecting id, activityDisplayName, initiatedBy, activityDateTime, result. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-directory-audit-events-list\n name: Audit directory audit events list\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_directory_audit_events.output | size }} directory audit events.\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-directory-audit-events-list, version 0.1.0, author Community/community) that reviews the tenant's directory audit events via /auditLogs/directoryAudits."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-directory-audit-events-list\n name: Review directory audit events list\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_directory_audit_events.output | size }} directory audit events.\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a compliance agent for OpenAdminOS: read directory audit events (fields: id, activityDisplayName, initiatedBy, activityDateTime, result; permission AuditLog.Read.All). Slug: check-directory-audit-events-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-directory-audit-events-list\n name: Check directory audit events list\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_directory_audit_events.output | size }} directory audit events.\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-directory-audit-events-list-scheduled` that loads directory audit events from Graph (/auditLogs/directoryAudits, GET, scope AuditLog.Read.All) selecting id, activityDisplayName, initiatedBy, activityDateTime, result. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-directory-audit-events-list-scheduled\n name: Check directory audit events list scheduled\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_directory_audit_events.output | size }} directory audit events.\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-directory-audit-events-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's directory audit events via /auditLogs/directoryAudits. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-directory-audit-events-list-scheduled\n name: Check directory audit events list scheduled\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_directory_audit_events.output | size }} directory audit events.\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a compliance agent for OpenAdminOS: read directory audit events (fields: id, activityDisplayName, initiatedBy, activityDateTime, result; permission AuditLog.Read.All). Slug: report-directory-audit-events-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-directory-audit-events-list-scheduled\n name: Report directory audit events list scheduled\n description: Reviews directory audit events and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: compliance\n mode: read\nskills:\n - id: load_directory_audit_events\n format: graph\n label: Load directory audit events\n detail: Reads directory audit events from the active tenant.\n settings:\n method: GET\n path: /auditLogs/directoryAudits\n select:\n - id\n - activityDisplayName\n - initiatedBy\n - activityDateTime\n - result\n scopes:\n - AuditLog.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_directory_audit_events.output | size }} directory audit events.\n data:\n total: '{{ load_directory_audit_events.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-service-principals-llm` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-service-principals-llm\n name: Report service principals llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-service-principals-llm, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-llm\n name: Review service principals llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read service principals (fields: id, displayName, appId, accountEnabled, servicePrincipalType; permission Application.Read.All), then have the LLM summarize what an admin should check. Slug: review-service-principals-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-llm\n name: Review service principals llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-service-principals-llm-scheduled` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-service-principals-llm-scheduled\n name: Check service principals llm scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-service-principals-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-service-principals-llm-scheduled\n name: Audit service principals llm scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-service-principals-count-llm` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType, breaks them down with a count transform, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-service-principals-count-llm\n name: Report service principals count llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: count_by_serviceprincipaltype\n format: transform\n label: Count service principals by servicePrincipalType\n detail: Breaks service principals down by servicePrincipalType.\n settings:\n kind: count-by-field\n source: '{{ load_service_principals.output }}'\n field: servicePrincipalType\n buckets:\n - Application\n - ManagedIdentity\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n count_by_serviceprincipaltype.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n breakdown: '{{ count_by_serviceprincipaltype.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-service-principals-count-llm, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-service-principals-count-llm\n name: Audit service principals count llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: count_by_serviceprincipaltype\n format: transform\n label: Count service principals by servicePrincipalType\n detail: Breaks service principals down by servicePrincipalType.\n settings:\n kind: count-by-field\n source: '{{ load_service_principals.output }}'\n field: servicePrincipalType\n buckets:\n - Application\n - ManagedIdentity\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n count_by_serviceprincipaltype.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n breakdown: '{{ count_by_serviceprincipaltype.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-service-principals-count-llm-scheduled` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType, breaks them down with a count transform, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-count-llm-scheduled\n name: Review service principals count llm scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: count_by_serviceprincipaltype\n format: transform\n label: Count service principals by servicePrincipalType\n detail: Breaks service principals down by servicePrincipalType.\n settings:\n kind: count-by-field\n source: '{{ load_service_principals.output }}'\n field: servicePrincipalType\n buckets:\n - Application\n - ManagedIdentity\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n count_by_serviceprincipaltype.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n breakdown: '{{ count_by_serviceprincipaltype.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-service-principals-count-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-service-principals-count-llm-scheduled\n name: Audit service principals count llm scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: count_by_serviceprincipaltype\n format: transform\n label: Count service principals by servicePrincipalType\n detail: Breaks service principals down by servicePrincipalType.\n settings:\n kind: count-by-field\n source: '{{ load_service_principals.output }}'\n field: servicePrincipalType\n buckets:\n - Application\n - ManagedIdentity\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n count_by_serviceprincipaltype.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n breakdown: '{{ count_by_serviceprincipaltype.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-service-principals-sort-llm` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType, samples the oldest by displayName, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-sort-llm\n name: Review service principals sort llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-service-principals-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-sort-llm\n name: Review service principals sort llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read service principals (fields: id, displayName, appId, accountEnabled, servicePrincipalType; permission Application.Read.All), show the oldest ten by displayName, then have the LLM summarize what an admin should check. Slug: review-service-principals-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-sort-llm\n name: Review service principals sort llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-service-principals-sort-llm-scheduled` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType, samples the oldest by displayName, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-service-principals-sort-llm-scheduled\n name: Check service principals sort llm scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-service-principals-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-sort-llm-scheduled\n name: Review service principals sort llm scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n load_service_principals.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `audit-service-principals-count-sort-llm` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType, breaks them down with a count transform, samples the oldest by displayName, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-service-principals-count-sort-llm\n name: Audit service principals count sort llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: count_by_serviceprincipaltype\n format: transform\n label: Count service principals by servicePrincipalType\n detail: Breaks service principals down by servicePrincipalType.\n settings:\n kind: count-by-field\n source: '{{ load_service_principals.output }}'\n field: servicePrincipalType\n buckets:\n - Application\n - ManagedIdentity\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n count_by_serviceprincipaltype.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n breakdown: '{{ count_by_serviceprincipaltype.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-service-principals-count-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals with a per-field breakdown and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-service-principals-count-sort-llm\n name: Check service principals count sort llm\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: count_by_serviceprincipaltype\n format: transform\n label: Count service principals by servicePrincipalType\n detail: Breaks service principals down by servicePrincipalType.\n settings:\n kind: count-by-field\n source: '{{ load_service_principals.output }}'\n field: servicePrincipalType\n buckets:\n - Application\n - ManagedIdentity\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n count_by_serviceprincipaltype.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n breakdown: '{{ count_by_serviceprincipaltype.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-service-principals-count-sort-llm-scheduled` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType, breaks them down with a count transform, samples the oldest by displayName, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-service-principals-count-sort-llm-scheduled\n name: Check service principals count sort llm scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: count_by_serviceprincipaltype\n format: transform\n label: Count service principals by servicePrincipalType\n detail: Breaks service principals down by servicePrincipalType.\n settings:\n kind: count-by-field\n source: '{{ load_service_principals.output }}'\n field: servicePrincipalType\n buckets:\n - Application\n - ManagedIdentity\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n count_by_serviceprincipaltype.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n breakdown: '{{ count_by_serviceprincipaltype.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-service-principals-count-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals with a per-field breakdown and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-service-principals-count-sort-llm-scheduled\n name: Report service principals count sort llm scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\n - id: count_by_serviceprincipaltype\n format: transform\n label: Count service principals by servicePrincipalType\n detail: Breaks service principals down by servicePrincipalType.\n settings:\n kind: count-by-field\n source: '{{ load_service_principals.output }}'\n field: servicePrincipalType\n buckets:\n - Application\n - ManagedIdentity\n - id: oldest_items\n format: transform\n label: Sample service principals by displayName\n detail: Lists service principals ordered by displayName.\n settings:\n kind: sort-by\n source: '{{ load_service_principals.output }}'\n field: displayName\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short service principals report.\n settings:\n system: >-\n You review service principals for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total service principals: {{ load_service_principals.output | size }}. Details: {{\n count_by_serviceprincipaltype.output }}. Write a compact report: main finding, what to\n verify, next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_service_principals.output | size }}'\n breakdown: '{{ count_by_serviceprincipaltype.output }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-service-principals-list` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-list\n name: Review service principals list\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_service_principals.output | size }} service principals.\n data:\n total: '{{ load_service_principals.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-service-principals-list, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-service-principals-list\n name: Audit service principals list\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_service_principals.output | size }} service principals.\n data:\n total: '{{ load_service_principals.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a policies agent for OpenAdminOS: read service principals (fields: id, displayName, appId, accountEnabled, servicePrincipalType; permission Application.Read.All). Slug: review-service-principals-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-list\n name: Review service principals list\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_service_principals.output | size }} service principals.\n data:\n total: '{{ load_service_principals.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-service-principals-list-scheduled` that loads service principals from Graph (/servicePrincipals, GET, scope Application.Read.All) selecting id, displayName, appId, accountEnabled, servicePrincipalType. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-service-principals-list-scheduled\n name: Check service principals list scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_service_principals.output | size }} service principals.\n data:\n total: '{{ load_service_principals.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-service-principals-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's service principals via /servicePrincipals. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-service-principals-list-scheduled\n name: Review service principals list scheduled\n description: Reviews service principals and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: policies\n mode: read\nskills:\n - id: load_service_principals\n format: graph\n label: Load service principals\n detail: Reads service principals from the active tenant.\n settings:\n method: GET\n path: /servicePrincipals\n select:\n - id\n - displayName\n - appId\n - accountEnabled\n - servicePrincipalType\n scopes:\n - Application.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_service_principals.output | size }} service principals.\n data:\n total: '{{ load_service_principals.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `review-intune-mobile-apps-llm` that loads Intune mobile apps from Graph (/deviceAppManagement/mobileApps, GET, scope DeviceManagementApps.Read.All) selecting id, displayName, publisher, createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-intune-mobile-apps-llm\n name: Review intune mobile apps llm\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-intune-mobile-apps-llm, version 0.1.0, author Community/community) that reviews the tenant's Intune mobile apps via /deviceAppManagement/mobileApps and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-intune-mobile-apps-llm\n name: Report intune mobile apps llm\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a apps agent for OpenAdminOS: read Intune mobile apps (fields: id, displayName, publisher, createdDateTime; permission DeviceManagementApps.Read.All), then have the LLM summarize what an admin should check. Slug: check-intune-mobile-apps-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-intune-mobile-apps-llm\n name: Check intune mobile apps llm\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-intune-mobile-apps-llm-scheduled` that loads Intune mobile apps from Graph (/deviceAppManagement/mobileApps, GET, scope DeviceManagementApps.Read.All) selecting id, displayName, publisher, createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-intune-mobile-apps-llm-scheduled\n name: Check intune mobile apps llm scheduled\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug review-intune-mobile-apps-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's Intune mobile apps via /deviceAppManagement/mobileApps and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: review-intune-mobile-apps-llm-scheduled\n name: Review intune mobile apps llm scheduled\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-intune-mobile-apps-sort-llm` that loads Intune mobile apps from Graph (/deviceAppManagement/mobileApps, GET, scope DeviceManagementApps.Read.All) selecting id, displayName, publisher, createdDateTime, samples the oldest by createdDateTime, and summarizes with an llm skill. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-intune-mobile-apps-sort-llm\n name: Check intune mobile apps sort llm\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: oldest_items\n format: transform\n label: Sample Intune mobile apps by createdDateTime\n detail: Lists Intune mobile apps ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_intune_mobile_apps.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-intune-mobile-apps-sort-llm, version 0.1.0, author Community/community) that reviews the tenant's Intune mobile apps via /deviceAppManagement/mobileApps and an LLM-written summary."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-intune-mobile-apps-sort-llm\n name: Audit intune mobile apps sort llm\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: oldest_items\n format: transform\n label: Sample Intune mobile apps by createdDateTime\n detail: Lists Intune mobile apps ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_intune_mobile_apps.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a apps agent for OpenAdminOS: read Intune mobile apps (fields: id, displayName, publisher, createdDateTime; permission DeviceManagementApps.Read.All), show the oldest ten by createdDateTime, then have the LLM summarize what an admin should check. Slug: check-intune-mobile-apps-sort-llm. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-intune-mobile-apps-sort-llm\n name: Check intune mobile apps sort llm\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: oldest_items\n format: transform\n label: Sample Intune mobile apps by createdDateTime\n detail: Lists Intune mobile apps ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_intune_mobile_apps.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-intune-mobile-apps-sort-llm-scheduled` that loads Intune mobile apps from Graph (/deviceAppManagement/mobileApps, GET, scope DeviceManagementApps.Read.All) selecting id, displayName, publisher, createdDateTime, samples the oldest by createdDateTime, and summarizes with an llm skill. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-intune-mobile-apps-sort-llm-scheduled\n name: Check intune mobile apps sort llm scheduled\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: oldest_items\n format: transform\n label: Sample Intune mobile apps by createdDateTime\n detail: Lists Intune mobile apps ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_intune_mobile_apps.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug audit-intune-mobile-apps-sort-llm-scheduled, version 0.1.0, author Community/community) that reviews the tenant's Intune mobile apps via /deviceAppManagement/mobileApps and an LLM-written summary. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: audit-intune-mobile-apps-sort-llm-scheduled\n name: Audit intune mobile apps sort llm scheduled\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\n - id: oldest_items\n format: transform\n label: Sample Intune mobile apps by createdDateTime\n detail: Lists Intune mobile apps ordered by createdDateTime.\n settings:\n kind: sort-by\n source: '{{ load_intune_mobile_apps.output }}'\n field: createdDateTime\n direction: asc\n take: 10\n - id: summarize\n format: llm\n label: Summarize findings with the active LLM\n detail: Generates a short Intune mobile apps report.\n settings:\n system: >-\n You review Intune mobile apps for Microsoft 365 admins. Be concise, factual, and\n conservative. Use only the supplied data.\n prompt: >-\n Total Intune mobile apps: {{ load_intune_mobile_apps.output | size }}. Details: {{\n load_intune_mobile_apps.output }}. Write a compact report: main finding, what to verify,\n next actions.\n temperature: 0.2\n maxTokens: 400\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: '{{ summarize.output.text | default(\"Summary unavailable.\") }}'\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n sample: '{{ oldest_items.output }}'\n llmSummary: '{{ summarize.output.text }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `check-intune-mobile-apps-list` that loads Intune mobile apps from Graph (/deviceAppManagement/mobileApps, GET, scope DeviceManagementApps.Read.All) selecting id, displayName, publisher, createdDateTime. It runs manually. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-intune-mobile-apps-list\n name: Check intune mobile apps list\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_intune_mobile_apps.output | size }} Intune mobile apps.\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug check-intune-mobile-apps-list, version 0.1.0, author Community/community) that reviews the tenant's Intune mobile apps via /deviceAppManagement/mobileApps."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-intune-mobile-apps-list\n name: Check intune mobile apps list\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_intune_mobile_apps.output | size }} Intune mobile apps.\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a apps agent for OpenAdminOS: read Intune mobile apps (fields: id, displayName, publisher, createdDateTime; permission DeviceManagementApps.Read.All). Slug: check-intune-mobile-apps-list. "},{"role":"assistant","content":"```yaml\ndescriptor:\n id: check-intune-mobile-apps-list\n name: Check intune mobile apps list\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n result:\n summary: Reviewed {{ load_intune_mobile_apps.output | size }} Intune mobile apps.\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Draft a read-mode agent manifest with slug `report-intune-mobile-apps-list-scheduled` that loads Intune mobile apps from Graph (/deviceAppManagement/mobileApps, GET, scope DeviceManagementApps.Read.All) selecting id, displayName, publisher, createdDateTime. It should run manually and on a schedule. Author is \"Community\" (handle \"community\"), version 0.1.0."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-intune-mobile-apps-list-scheduled\n name: Report intune mobile apps list scheduled\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 86400\n result:\n summary: Reviewed {{ load_intune_mobile_apps.output | size }} Intune mobile apps.\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. An agent manifest is YAML with three required top-level keys: descriptor, skills, and definition. descriptor requires: id (kebab-case slug), name, description, version (semver), minAppVersion, author {name, handle}, category, mode (read or write). Each skill has: id (snake_case, no hyphens), format (graph | transform | llm), label, detail, settings. Graph settings: method, path, select, scopes. Transform settings: kind (count-by-field | sort-by | group-by-field | group-by-age | filter-by-age), source \"{{ .output }}\", and kind-specific fields. LLM settings: system, prompt, temperature, maxTokens. definition holds triggers ({id, kind: manual|scheduled, intervalSeconds}) and result (summary + data map). Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"Create an OpenAdminOS agent (slug report-intune-mobile-apps-list-scheduled, version 0.1.0, author Community/community) that reviews the tenant's Intune mobile apps via /deviceAppManagement/mobileApps. Include both a manual and a scheduled trigger."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-intune-mobile-apps-list-scheduled\n name: Report intune mobile apps list scheduled\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_intune_mobile_apps.output | size }} Intune mobile apps.\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n```"}]} +{"messages":[{"role":"system","content":"You draft OpenAdminOS Agent Template manifests. Reply with exactly one fenced yaml code block containing only the manifest."},{"role":"user","content":"I need a apps agent for OpenAdminOS: read Intune mobile apps (fields: id, displayName, publisher, createdDateTime; permission DeviceManagementApps.Read.All). Slug: report-intune-mobile-apps-list-scheduled. Schedule it to recur."},{"role":"assistant","content":"```yaml\ndescriptor:\n id: report-intune-mobile-apps-list-scheduled\n name: Report intune mobile apps list scheduled\n description: Reviews Intune mobile apps and reports findings for admin review.\n version: 0.1.0\n minAppVersion: 0.1.0\n author:\n name: Community\n handle: community\n category: apps\n mode: read\nskills:\n - id: load_intune_mobile_apps\n format: graph\n label: Load Intune mobile apps\n detail: Reads Intune mobile apps from the active tenant.\n settings:\n method: GET\n path: /deviceAppManagement/mobileApps\n select:\n - id\n - displayName\n - publisher\n - createdDateTime\n scopes:\n - DeviceManagementApps.Read.All\ndefinition:\n triggers:\n - id: manual\n kind: manual\n - id: schedule\n kind: scheduled\n intervalSeconds: 604800\n result:\n summary: Reviewed {{ load_intune_mobile_apps.output | size }} Intune mobile apps.\n data:\n total: '{{ load_intune_mobile_apps.output | size }}'\n```"}]}