Apple II Universal Machine — User Guide
A complete guide to operating the browser-native agent OS cockpit.
Table of Contents
- Getting Started
- Terminal Basics
- Core Commands
- Ada Contracts
- Lisp Machine
- Assembly & Encoding
- Heterogeneous VM Lab
- Fontana → Lisp → VM Pipeline
- Agent System
- Digital Twins
- Mode Switching
- Ollama Agent-in-a-Box
- Vacuum Pipeline
- Envelope System
- Export & Screenshots
- Repo Staple & Cold Boot
- Capsule System
- Model Architectures
- GitHub-Native Integration
- Keyboard Shortcuts
- Troubleshooting
- FAQ
Getting Started
Opening the Cockpit
- Navigate to
https://SNAPKITTYWEST.github.io/apple-ii-universal-machine/ - Watch the boot sequence (ASCII art + status messages)
- The terminal appears with
HolySim>prompt - Type
Help;to see all commands
Two Terminals
- index.html — Main terminal with status panel, audit log, runtime bar
- holy-terminal.html — Dedicated HolyC terminal with 100+ commands, button bar, full status
Both share the same Woz Vault (localStorage) and seal engine.
First Commands
Help; Show all commands
Boot; Reboot with ASCII art
Status; System status + runtime detection
Seal; Generate SHA-256 seal
Debate("consciousness"); Run ENKI/SENTINEL debate
Terminal Basics
Input Format
Commands use HolyC-style syntax with semicolons:
CommandName;
CommandName "argument";
CommandName "arg1" "arg2";
Or parenthetical style:
CommandName(argument);
Command History
- ↑ Arrow — Previous command
- ↓ Arrow — Next command
- Enter — Execute
Button Bar
Click buttons for quick access to common commands. Buttons are grouped by layer.
Status Bar
Shows live state:
VAULT: N— Woz Vault event countSEALS: N— Total seals generatedOLLAMA: ONLINE/OFFLINE— Ollama connectionMODEL: llama3.2— Current modelMODE: APPLE— Active modeAGENTS: N— Registered agentsTWINS: N— Registered twinsCAPSULES: N— Registered capsulesREPO STAPLE: VERIFIED/UNVERIFIED— Staple statusGITHUB NATIVE: READY/AUTH REQUIRED— GitHub auth
Core Commands
Help
Help;
Displays the full command reference organized by layer.
Boot
Boot;
Runs Ada contract check, displays ASCII art boot sequence, generates seal, writes audit event.
Status
Status;
Shows:
- System version
- Vault event count
- Seal count
- Runtime status (all adapters)
- Agent/twin/capsule counts
Trust
Trust;
Runs the 8-gate trust check:
local_only— No remote executionnon_destructive— No data destructionauditable— All actions loggedsealable— SHA-256 availablerecoverable— State can be restored
Seal
Seal;
Seal "custom payload";
Generates SHA-256 hash via Web Crypto API. Writes to Woz Vault.
Woz Vault
Vault;
Displays the full audit log with timestamps, event types, and seal hashes.
Clear Vault
ClearVault;
Wipes all Woz Vault data. Requires Ada contract confirmation.
Export
Export;
Downloads complete audit log + snapshots as JSON file.
About
About;
System information including runtime, agents, contracts, and philosophy.
Ada Contracts
Ada Status
AdaStatus;
Shows all 7 contract gates and their status.
Contracts
Contracts;
Lists all contract gates with descriptions.
How Contracts Work
Ada contracts are real .ads/.adb source files enforced by a JavaScript simulator:
Boot— Requires user confirmationDebate— Always allowedWriteAudit— Always allowedExportSnapshot— Always allowedClearVault— Requires user confirmationInvokeLisp— Always allowedInvokeRuntime— Always allowed
When a command requires a contract check, it runs through ContractRegistry.check() before execution.
Lisp Machine
Lisp Help
LispHelp;
Shows all Lisp commands.
Evaluate Expressions
LispEval (+ 1 2);
LispEval (* 10 5);
LispEval (- 100 37);
Arithmetic operations on s-expressions.
System Queries
LispEval (agent-status);
LispEval (trust-score);
LispEval (woz-vault-count);
LispEval (seal-state);
Query system state through Lisp.
ASCII Art
LispEval (draw-agent);
Generate ASCII art representation of agents.
Fontana Reaction
FontanaReact (agent trust audit);
FontanaReact (quantum consciousness);
Symbolic reaction chamber — transforms input symbols through reaction rules.
Assembly & Encoding
Assembly Status
AsmStatus;
Shows assembly layer status (simulated register/opcode viewer).
CPU Registers
Registers;
Displays current CPU register state:
AX— AccumulatorBX— BaseCX— CounterDX— DataIP— Instruction PointerFLAGS— Status flags
Opcodes
Opcodes Boot;
Opcodes Seal;
Opcodes Deploy;
Shows the opcode translation for a given command.
Reverse Unicode
ReverseUnicode "Hello World";
Reverses and mirrors a string using Unicode character mapping.
Glyph Seal
GlyphSeal "No token tax";
Creates a visual seal with:
- Original text
- Reversed text
- Mirrored text
- SHA-256 hash
Restore Unicode
RestoreUnicode "dlroW olleH";
Restores a mirrored/reversed string to original.
Heterogeneous VM Lab
Prolog Gate
PrologAsk allowed(boot_agent);
PrologAsk requires_seal(deploy_code);
Symbolic rule lookup against 18 predefined rules. Returns ALLOWED, DENIED, or REQUIRES_SEAL.
Macro Expansion
MacroExpand boot_agent;
MacroExpand deploy_code;
MacroExpand audit_trail;
Expands high-level intent into step-by-step execution plan.
Bytecode VM
BytecodeRun PUSH 1 PUSH 2 ADD PRINT;
BytecodeRun PUSH 5 PUSH 3 SUB PRINT;
Safe stack-based interpreter with operations:
PUSH value— Push to stackADD— Add top twoSUB— SubtractMUL— MultiplyPRINT— Output topHALT— Stop
1000 step limit. No external dependencies.
Brainfuck Tape
BrainfuckRun ++>+++<[->+<]>.;
Sandboxed Brainfuck interpreter:
- 30,000 cells
- 5,000 step limit
- Bracket balance validation
- No infinite loop risk
Machine Viewer
MachineView boot_agent;
MachineView deploy_code;
Generates fake opcode artifact for visual representation. NOT EXECUTED.
VM Trace
VMTrace boot_agent;
VMTrace audit_trail;
Runs full pipeline: Prolog → Macros → Bytecode → Machine View → SHA-256 seal.
Fontana → Lisp → VM Pipeline
Fontana Decode
FontanaDecode agent trust audit;
FontanaDecode deploy payment received;
Decodes Fontana symbols into:
- Lisp s-expression
- Bytecode operations
Lisp Expand
LispExpand (agent trust audit);
LispExpand (deploy payment received);
Expands a Lisp expression into detailed form.
Lisp to VM
LispToVM (boot);
LispToVM (agent trust audit);
Compiles Lisp expression to VM bytecode.
Full Pipeline Example
FontanaDecode agent trust audit;
→ Produces Lisp: (event agent_trust_audit)
→ Produces bytecode: [PUSH_EVENT, DEFINE_EVENT, LOAD_AGENT, CHECK_TRUST, SEAL, HALT]
Agent System
Agent Help
AgentHelp;
Shows all agent commands.
Spawn Agent
SpawnAgent "Researcher";
SpawnAgent "Auditor" auditor;
SpawnAgent "Builder" builder;
Creates a new agent with:
- Unique ID (
agent_...) - Name
- Role (8 options)
- Tools
- Permissions
- Memory scope
- SHA-256 seal
Agent Roles
| Role | Tools | Permissions |
|---|---|---|
researcher |
search, summarize, cite | read, audit |
auditor |
verify, seal, log | read, audit |
builder |
create, edit, compile | read, write |
debater |
propose, challenge, synthesize | read, audit |
archivist |
store, index, retrieve | read, write |
sentinel |
monitor, alert, block | read, audit |
explorer |
scan, map, discover | read |
translator |
decode, encode, transform | read, write |
List Agents
ListAgents;
Shows all registered agents with run counts.
Run Agent
RunAgent "Researcher" "analyze system";
RunAgent "Auditor" "verify trust deed";
Runs agent on task. Generates role-appropriate response. Writes audit event + seal.
Inspect Agent
InspectAgent "Researcher";
Shows full agent profile: ID, name, role, tools, permissions, memory, mode, runs, seal.
Delete Agent
DeleteAgent "Researcher";
Removes agent from registry.
Digital Twins
Twin Help
TwinHelp;
Shows all twin commands.
Create Twin
CreateTwin "Analyst";
Creates a digital twin with:
- Unique ID (
twin_...) - Name
- Default role: analyst
- Default personality: methodical
- SHA-256 seal
Design Twin
DesignTwin "Analyst" role="researcher" memory="persistent";
DesignTwin "Analyst" personality="methodical" permissions="read,audit";
Customize twin specifications.
Twin Roles
analyst— Pattern analysis, data flow, risk assessmentbuilder— Component identification, build planningresearcher— Source analysis, findings summary
List / Run / Inspect / Delete
ListTwins;
RunTwin "Analyst" "review code quality";
InspectTwin "Analyst";
DeleteTwin "Analyst";
Same pattern as agents.
Mode Switching
List Modes
ListModes;
Shows all 6 modes with themes:
| Mode | Theme | Focus |
|---|---|---|
apple |
terminal-green | Full command set |
linux |
terminal-amber | Shell-focused |
windows |
terminal-blue | PowerShell-style |
holy |
terminal-gold | Faith-governed |
vm |
terminal-cyan | VM-focused |
agent |
terminal-magenta | Agent-focused |
Switch Mode
SwitchMode linux;
SwitchMode vm;
SwitchMode holy;
Filters available commands. Updates prompt and theme.
Current Mode
CurrentMode;
Shows active mode name, theme, and prompt.
Ollama Agent-in-a-Box
Ollama Status
OllamaStatus;
Shows connection status, URL, model, model count, last probe time.
Connect Ollama
ConnectOllama;
Probes http://localhost:11434 with 3-second timeout.
If offline: OLLAMA: OFFLINE — using simulated agent mode
If online: Lists detected models.
List Models
ListModels;
Shows 10 default models + any detected local models:
| Model | Best For |
|---|---|
llama3.2 |
General purpose |
llama3.1 |
Fast inference |
mistral |
Reasoning |
codellama |
Code generation |
phi3 |
Compact |
gemma2 |
Balanced |
qwen2.5 |
Multilingual |
deepseek-coder |
Code specialist |
neural-chat |
Conversational |
starling-lm |
Alignment |
Spawn Ollama Agent
SpawnAgent "Researcher";
Creates agent with current Ollama model.
Design Agent
DesignAgent "Researcher" role="research" model="llama3.2" mode="linux";
Configure agent model and role.
Run Agent
RunAgent "Researcher" "Explain local-first AI";
If Ollama is online: sends to localhost:11434/api/generate
If offline: returns simulated response with label
Vacuum Pipeline
Vacuum Help
VacuumHelp;
Shows vacuum commands and pipeline.
Vacuum Ingest
VacuumIngest "The customer paid the invoice on Tuesday";
Stores raw text with:
- Word count
- Sentence count
- Character count
- Timestamp
Extract Meaning
ExtractMeaning "The customer paid the invoice on Tuesday";
Extracts:
- Entities (customer, invoice)
- Actions (paid)
- Topics (payment, customer)
- Sentiment (-1 to +1)
- Word/sentence counts
To Lisp
ToLisp "customer paid invoice";
ToLisp "deploy payment received";
Full pipeline:
- Extract meaning from text
- Map to normalized schema
- Generate Lisp s-expression
Output:
EVENT: customer_paid_invoice
CATEGORY: payment
LISP: (pipeline (event customer_paid_invoice) (schema (category payment) (confidence 0.7)))
Envelope System
Compile Envelope
CompileEnvelope "(event payment received)";
CompileEnvelope "(event agent_trust_audit)";
Pipeline:
- Parse Lisp expression
- Compile to bytecode
- Build signed envelope
- Execute bytecode trace
- Write audit event + seal
Envelope format:
{
"envelopeId": "env_...",
"type": "BYTECODE_ENVELOPE",
"source": "(event payment received)",
"lisp": "(event payment received)",
"bytecode": ["PUSH_EVENT", "DEFINE_EVENT", "VERIFY", "SEAL", "HALT"],
"trust": "ALLOWED",
"seal": "sha256...",
"createdAt": "2026-06-17T...",
"mode": "apple",
"verified": true
}
Verify Envelope
VerifyEnvelope;
8-check verification:
- Has ID
- Has type
- Has source
- Has Lisp expression
- Has bytecode array
- Has seal
- Has trust
- Seal length ≥ 32
Run Envelope
RunEnvelope;
Executes verified bytecode. Outputs step trace. Writes audit + seal.
Export Envelope
ExportEnvelope;
Downloads envelope as JSON file.
Export & Screenshots
Screenshot Terminal
ScreenshotTerminal;
Renders terminal content to canvas, downloads as PNG image.
Export Proof
ExportProof;
Full proof package including:
- Agent name
- Model
- Prompt
- Response
- Timestamp
- Mode
- Seal
- Audit count
Downloads as JSON.
Repo Staple & Cold Boot
Staple Repo
StapleRepo;
Generates manifest:
{
"repo": "apple-ii-universal-machine",
"version": "1.0.0",
"files": [...],
"fileCount": 100,
"buildTimestamp": "...",
"pageSeal": "sha256...",
"vaultSnapshotHash": "sha256...",
"vaultEvents": 42,
"agentCount": 3,
"twinCount": 2,
"capsuleCount": 1,
"mode": "apple",
"verified": true
}
Verify Staple
VerifyStaple;
Checks 7 conditions:
- Has repo name
- Has version
- Has files
- Has timestamp
- Has page seal
- Has vault hash
- Seal length ≥ 32
Cold Boot
ColdBoot;
Replays boot sequence:
- Read staple manifest
- Check Woz Vault
- Verify page seal
- Restore last known state
- Print SUN BOOT COMPLETE
Staple Status
StapleStatus;
Shows verified/unverified, repo, version, file count, page seal, cold boot status.
Capsule System
Capsule Help
CapsuleHelp;
Shows all capsule commands.
Create Capsule
CreateCapsule "Researcher";
Creates capsule with:
- Name
- Type: agent
- Architecture: transformer
- Model: ollama/llama3.2
- Runtime: ollama
- Tools, permissions, evals
- Limitations (honest labels)
- Intended use
- SHA-256 seal
Design Capsule
DesignCapsule "Researcher" arch="mamba" model="ollama/mistral" runtime="ollama";
DesignCapsule "Researcher" limitations="Simulated only|No remote calls" intendedUse="Local research";
List Capsules
ListCapsules;
Shows all capsules with type, architecture, and model.
Inspect Capsule
InspectCapsule "Researcher";
Full capsule profile.
Run Capsule
RunCapsule "Researcher" "Explain local-first AI";
If Ollama online: sends to model If offline: returns simulated response
Verify Capsule
VerifyCapsule "Researcher";
8-check verification of capsule metadata.
Staple Capsule
StapleCapsule "Researcher";
Links capsule to repo staple manifest.
Export Capsule
ExportCapsule "Researcher";
Downloads capsule as JSON.
Generate Agent Card
GenerateAgentCard "Researcher";
Downloads Markdown card with:
- Intended use
- Limitations
- Model/runtime
- Memory policy
- Permissions
- Eval commands
- Woz Vault seal
- Repo staple info
Generate Model Card
GenerateModelCard "Researcher";
Downloads model card with architecture, base model, training status (SIMULATED).
Model Architectures
List Architectures
ListArchitectures;
Shows available architectures:
| Architecture | Description |
|---|---|
transformer |
Attention-based, standard LLM |
mamba |
State-space model, linear complexity |
hybrid |
Transformer + Mamba combination |
Select Architecture
When creating or designing capsules:
CreateCapsule "Researcher" arch="mamba";
DesignCapsule "Researcher" arch="hybrid";
GitHub-Native Integration
GitHub Help
GitHubHelp;
Shows all GitHub commands.
GitHub Login
GitHubLogin "ghp_your_token_here";
Security rules:
- Token stored in
sessionStorageonly (notlocalStorage) - Never committed to repo
- Never written to Woz Vault
- Never shown in terminal output
- Cleared on page close
GitHub Status
GitHubStatus;
Shows authentication status, username, scopes.
GitHub Logout
GitHubLogout;
Clears token from sessionStorage.
Create Inverted Repo
CreateInvertedRepo "my-agent-os";
CreateInvertedRepo "my-twin-os" twin;
CreateInvertedRepo "my-vacuum-os" vacuum;
CreateInvertedRepo "my-vm-lab" vm;
Creates GitHub repo via API with:
- README.md
- TRUST_DEED.md
- capsule.json
- woz-staple.json
- index.html
- .github/workflows/pages.yml
- .github/workflows/capsule-verify.yml
Create Repo From Template
CreateRepoFromTemplate "my-agent-os";
Same as CreateInvertedRepo with template selection.
Staple Remote Repo
StapleRemoteRepo "owner/repo";
Writes woz-staple.json to remote repo.
Verify Remote Repo
VerifyRemoteRepo "owner/repo";
Reads and verifies remote staple manifest.
Keyboard Shortcuts
| Key | Action |
|---|---|
Enter |
Execute command |
↑ |
Previous command |
↓ |
Next command |
Tab |
(not yet implemented) |
Troubleshooting
"Unknown: command"
Command not recognized. Type Help; for full list.
"ADA CONTRACT: DENIED"
Command requires contract confirmation. For ClearVault;, confirm when prompted.
"Ollama: OFFLINE"
Ollama not running at localhost:11434. Start Ollama or use simulated mode.
"No envelope built"
Run CompileEnvelope before VerifyEnvelope or RunEnvelope.
"Agent not found"
Agent name must match exactly (case-sensitive). Run ListAgents; to see registered agents.
"Token not valid"
GitHub token must start with ghp_ or github_pat_. Check token permissions.
Vault seems empty
Run Boot; or Seal; to generate events. Check Vault; for audit log.
FAQ
Is this real HolyC?
No. This is a browser-safe simulation. All execution is JavaScript in the browser. The Ada contract source files (.ads/.adb) are real Ada but enforced by a JS simulator.
Does Ollama work?
If you have Ollama running at localhost:11434, yes. Otherwise, all agent runs use simulated mode with honest labels.
Can I break it?
No. All VM execution is sandboxed (step limits, cell limits). Brainfuck has 5000 steps max. Bytecode VM has 1000 steps max.
Where is data stored?
All data is in browser localStorage. No server. No database. No cloud. Clear your browser data to reset.
Can I export everything?
Yes. Export; downloads audit log. ExportProof; downloads full proof package. ExportCapsule "N"; downloads capsule JSON. ScreenshotTerminal; downloads terminal as PNG.
How do I verify integrity?
Run VerifyStaple; to check repo staple. Run VerifyCapsule "N"; to check capsule. Run Trust; to check trust gates. Every action generates a SHA-256 seal.
What is cold boot?
ColdBoot; reads the repo staple manifest, checks Woz Vault, verifies the page seal, restores last known state, and prints SUN BOOT COMPLETE. It proves the live page is attached to the repo state.
What is an inverted repo?
An inverted repo is where the source of truth is architectural intent and trust rules, not applications. The repo IS the runtime manifest, proof source, capsule source, and cold-boot source.
Apple II Universal Machine · Browser-native agent OS cockpit · No token tax