Add files using upload-large-folder tool

tasks/format-code-task-000034/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:97cb9a9081323508cf4eaa282765cfe9253530e431d20cd5e74758bc96193edb

tasks/format-code-task-000034/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com

tasks/format-code-task-000034/environment/setup/setup.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# Runs once, as root, before the agent starts: Harbor's [environment.healthcheck] unpacks this folder into
+# /var/lib/mimo and runs it. Generated by mimo_harbor; see its README for how each step maps to Xiaomi's harness.
+M=/var/lib/mimo
+[ -f "$M/ready" ] && exit 0
+chmod 700 "$M"
+fail() { echo "setup: $*" >&2; exit 1; }
+# the answer-leak blocklist (mimoagent), applied to /etc/hosts by the agent right after it installs
+write_blocklist() { cp "$M/files/blocklist" "$M/blocklist"; }
+
+# Code: mimoagent environments/datasets/opensource_code.py + base.py, as the explorer runs it (app/runner/domains.py Code).
+CWD=/testbed
+git config --global --add safe.directory "$CWD"
+cd "$CWD" || fail "no working directory $CWD"
+if ! git rev-parse --git-dir >/dev/null 2>&1; then   # a bare source tree: give it a baseline commit
+  git init -q && git add -A && git -c user.name=mimo -c user.email=mimo@localhost commit -q -m baseline --allow-empty
+fi
+BASE=$(git rev-parse HEAD 2>/dev/null)
+[ ${#BASE} -eq 40 ] || fail "could not resolve the base commit in $CWD"
+echo "$BASE" > "$M/base"
+# Images are built with history truncated at the base. If one is not, the fix could be read out of git log,
+# so .git is hidden while the agent works and put back for grading.
+LATER=$(git rev-list --all --not "$BASE" 2>/dev/null | head -n 5 | grep -c . || true)
+rm -f /tmp/fail.log /tmp/pass.log /tmp/patch.diff /tmp/test_patch.diff /tmp/*.log 2>/dev/null; rm -rf /tmp/claude-0 /tmp/claude-* /tmp/testem-* /tmp/puppeteer_dev_chrome_profile-* 2>/dev/null; rm -f /tmp/test_files.json 2>/dev/null; rm -rf /tmp/jest_* /tmp/jest-* /tmp/build /tmp/build_env 2>/dev/null; rm -rf /tmp/pytest-of-root /tmp/pytest-* /tmp/.pytest_cache /tmp/__pycache__ 2>/dev/null; rm -rf /tmp/go-build* /tmp/standards /tmp/wordpress /tmp/zig-* /tmp/testbase /tmp/testbed 2>/dev/null; rm -f /tmp/tmp*.tmp /tmp/*.bak /tmp/expect* /tmp/butwas* 2>/dev/null; rm -rf /root/.cache/go-build 2>/dev/null; rm -rf /tests 2>/dev/null; find /logs -mindepth 1 -delete 2>/dev/null; mkdir -p /logs/agent /logs/verifier /logs/artifacts; chmod 777 /logs/agent /logs/verifier /logs/artifacts; find /var/log -type f -delete 2>/dev/null || true; rm -rf /var/lib/postgresql/*/*/log /var/lib/postgresql/*/*/pg_log 2>/dev/null || true; find /var/lib/mysql /var/lib/mongodb -type f -name '*.log' -delete 2>/dev/null || true; rm -rf /root/.npm/_logs /root/.babel.json /root/.pytest_cache 2>/dev/null || true; find / -maxdepth 4 -xdev -name task_description.md -path '*/.build_env/*' -delete 2>/dev/null || true; true
+git clean -fdx --exclude=node_modules --exclude=bower_components --exclude=.husky --exclude=vendor --exclude=third_party --exclude=_deps --exclude=vcpkg_installed --exclude=.venv --exclude=venv --exclude=.gradle --exclude=target --exclude=.build --exclude=lib --exclude=.bundle --exclude=Manifest.toml --exclude=_build --exclude=.stack-work --exclude=dist-newstyle >/dev/null 2>&1
+find /root/.m2 -type f \( -name '*-SNAPSHOT.jar' -o -name '*-SNAPSHOT-sources.jar' -o -name '*-SNAPSHOT-tests.jar' -o -name '*-SNAPSHOT-test-sources.jar' \) -delete 2>/dev/null; rm -rf /root/.julia/compiled 2>/dev/null; rm -rf /root/.gradle/caches/build-cache-* /root/.gradle/daemon 2>/dev/null; rm -rf /root/.gradle/caches/*/scripts /root/.gradle/caches/jars-* 2>/dev/null; rm -rf /root/.cache/bazel 2>/dev/null; true
+if [ "$LATER" -gt 0 ]; then
+  mv "$CWD/.git" "$M/git-hidden"
+  echo "history not truncated at ${BASE:0:12}: .git hidden while the agent works"
+fi
+write_blocklist
+
+touch "$M/ready"
+echo "setup done"

tasks/format-code-task-000034/tests/test.patch

@@ -0,0 +1,259 @@
+diff --git a/cli/azd/internal/vsrpc/handler_test.go b/cli/azd/internal/vsrpc/handler_test.go
+index a4f33297f..9d78cc34d 100644
+--- a/cli/azd/internal/vsrpc/handler_test.go
++++ b/cli/azd/internal/vsrpc/handler_test.go
+@@ -3,6 +3,7 @@ package vsrpc
+ import (
+ 	"context"
+ 	"errors"
++	"strings"
+ 	"testing"
+ 
+ 	"github.com/stretchr/testify/require"
+@@ -83,6 +84,144 @@ func TestHandler(t *testing.T) {
+ 	t.Run("HandlerFunc3", func(t *testing.T) { runHandlerSuite(t, newHandlerFunc3, []any{"arg0", "arg1", "arg2"}, "ok") })
+ }
+ 
++type panicHandlerCase struct {
++	name    string
++	message string
++	handler Handler
++	request jsonrpc2.Request
++}
++
++func TestPanicRecoveredAcrossHandlerAdapters(t *testing.T) {
++	cases := []panicHandlerCase{
++		{
++			name:    "action zero args",
++			message: "hidden action zero panic",
++			handler: HandlerAction0(func(context.Context) error {
++				panicRpcTestFrame("hidden action zero panic")
++				return nil
++			}),
++			request: requestWithParams(t),
++		},
++		{
++			name:    "action one arg",
++			message: "hidden action one panic",
++			handler: HandlerAction1(func(ctx context.Context, value string) error {
++				require.Equal(t, "alpha", value)
++				panicRpcTestFrame("hidden action one panic")
++				return nil
++			}),
++			request: requestWithParams(t, "alpha"),
++		},
++		{
++			name:    "action two args",
++			message: "hidden action two panic",
++			handler: HandlerAction2(func(ctx context.Context, value string, count int) error {
++				require.Equal(t, "bravo", value)
++				require.Equal(t, 42, count)
++				panicRpcTestFrame("hidden action two panic")
++				return nil
++			}),
++			request: requestWithParams(t, "bravo", 42),
++		},
++		{
++			name:    "action three args",
++			message: "hidden action three panic",
++			handler: HandlerAction3(func(ctx context.Context, value string, count int, flag bool) error {
++				require.Equal(t, "charlie", value)
++				require.Equal(t, 7, count)
++				require.True(t, flag)
++				panicRpcTestFrame("hidden action three panic")
++				return nil
++			}),
++			request: requestWithParams(t, "charlie", 7, true),
++		},
++		{
++			name:    "func zero args",
++			message: "hidden func zero panic",
++			handler: HandlerFunc0(func(context.Context) (string, error) {
++				panicRpcTestFrame("hidden func zero panic")
++				return "", nil
++			}),
++			request: requestWithParams(t),
++		},
++		{
++			name:    "func one arg",
++			message: "hidden func one panic",
++			handler: HandlerFunc1(func(ctx context.Context, value string) (string, error) {
++				require.Equal(t, "delta", value)
++				panicRpcTestFrame("hidden func one panic")
++				return "", nil
++			}),
++			request: requestWithParams(t, "delta"),
++		},
++		{
++			name:    "func two args",
++			message: "hidden func two panic",
++			handler: HandlerFunc2(func(ctx context.Context, value string, count int) (string, error) {
++				require.Equal(t, "echo", value)
++				require.Equal(t, 84, count)
++				panicRpcTestFrame("hidden func two panic")
++				return "", nil
++			}),
++			request: requestWithParams(t, "echo", 84),
++		},
++		{
++			name:    "func three args",
++			message: "hidden func three panic",
++			handler: HandlerFunc3(func(ctx context.Context, value string, count int, flag bool) (string, error) {
++				require.Equal(t, "foxtrot", value)
++				require.Equal(t, 21, count)
++				require.False(t, flag)
++				panicRpcTestFrame("hidden func three panic")
++				return "", nil
++			}),
++			request: requestWithParams(t, "foxtrot", 21, false),
++		},
++	}
++
++	for _, tc := range cases {
++		t.Run(tc.name, func(t *testing.T) {
++			var repliedErr error
++			reply := func(ctx context.Context, result any, err error) error {
++				repliedErr = err
++				return err
++			}
++
++			var handlerErr error
++			require.NotPanics(t, func() {
++				handlerErr = tc.handler(context.Background(), nil, reply, tc.request)
++			})
++
++			require.Error(t, handlerErr)
++			require.Same(t, repliedErr, handlerErr)
++			assertPanicRpcError(t, handlerErr, tc.message)
++		})
++	}
++}
++
++func requestWithParams(t *testing.T, params ...any) jsonrpc2.Request {
++	t.Helper()
++
++	req, err := jsonrpc2.NewCall(jsonrpc2.NewNumberID(1), "hidden/test", params)
++	require.NoError(t, err)
++
++	return req
++}
++
++func assertPanicRpcError(t *testing.T, err error, panicMessage string) {
++	t.Helper()
++
++	var rpcErr *jsonrpc2.Error
++	require.True(t, errors.As(err, &rpcErr), "expected JSON-RPC error, got %T: %v", err, err)
++	require.Equal(t, jsonrpc2.InternalError, rpcErr.Code)
++	require.True(t, strings.HasPrefix(rpcErr.Message, "panic:"), "message should start with panic:, got %q", rpcErr.Message)
++	require.Contains(t, rpcErr.Message, panicMessage)
++}
++
++func panicRpcTestFrame(message string) {
++	panic(message)
++}
++
+ func newHandlerAction0(t *testing.T, tc handlerTestCase) Handler {
+ 	return HandlerAction0(func(ctx context.Context) error {
+ 		select {
+diff --git a/cli/azd/internal/vsrpc/server_test.go b/cli/azd/internal/vsrpc/server_test.go
+index 4ed8aad0e..3f5a3016c 100644
+--- a/cli/azd/internal/vsrpc/server_test.go
++++ b/cli/azd/internal/vsrpc/server_test.go
+@@ -6,8 +6,10 @@ import (
+ 	"errors"
+ 	"net/http/httptest"
+ 	"net/url"
++	"strings"
+ 	"sync"
+ 	"testing"
++	"time"
+ 
+ 	"github.com/gorilla/websocket"
+ 	"github.com/stretchr/testify/require"
+@@ -132,3 +134,64 @@ func TestObserverable(t *testing.T) {
+ 	// The onCompleted message takes no parameters and the args value is empty.
+ 	require.Len(t, onCompletedParams[0], 0)
+ }
++
++func TestDebugServicePanicErrorDetails(t *testing.T) {
++	rpcConn := debugRpcConn(t)
++
++	err := callTestPanicAsync(t, rpcConn, "panic probe alpha hidden")
++	assertDebugPanicError(t, err, "panic probe alpha hidden")
++}
++
++func TestDebugServicePanicDoesNotCloseConnection(t *testing.T) {
++	rpcConn := debugRpcConn(t)
++
++	firstErr := callTestPanicAsync(t, rpcConn, "panic probe beta one hidden")
++	assertDebugPanicError(t, firstErr, "panic probe beta one hidden")
++
++	secondErr := callTestPanicAsync(t, rpcConn, "panic probe beta two hidden")
++	assertDebugPanicError(t, secondErr, "panic probe beta two hidden")
++}
++
++func debugRpcConn(t *testing.T) jsonrpc2.Conn {
++	t.Helper()
++
++	debugServer := httptest.NewServer(newDebugService())
++	t.Cleanup(debugServer.Close)
++
++	serverUrl, err := url.Parse(debugServer.URL)
++	require.NoError(t, err)
++	serverUrl.Scheme = "ws"
++
++	wsConn, _, err := websocket.DefaultDialer.Dial(serverUrl.String(), nil)
++	require.NoError(t, err)
++	t.Cleanup(func() {
++		require.NoError(t, wsConn.Close())
++	})
++
++	rpcConn := jsonrpc2.NewConn(newWebSocketStream(wsConn))
++	rpcConn.Go(context.Background(), nil)
++
++	return rpcConn
++}
++
++func callTestPanicAsync(t *testing.T, rpcConn jsonrpc2.Conn, message string) error {
++	t.Helper()
++
++	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
++	defer cancel()
++
++	_, err := rpcConn.Call(ctx, "TestPanicAsync", []any{message}, nil)
++	require.Error(t, err)
++
++	return err
++}
++
++func assertDebugPanicError(t *testing.T, err error, panicMessage string) {
++	t.Helper()
++
++	var rpcErr *jsonrpc2.Error
++	require.True(t, errors.As(err, &rpcErr), "expected JSON-RPC error, got %T: %v", err, err)
++	require.Equal(t, jsonrpc2.InternalError, rpcErr.Code)
++	require.True(t, strings.HasPrefix(rpcErr.Message, "panic:"), "message should start with panic:, got %q", rpcErr.Message)
++	require.Contains(t, rpcErr.Message, panicMessage)
++}
+diff --git a/mimo_test_command.sh b/mimo_test_command.sh
+new file mode 100755
+--- /dev/null
++++ b/mimo_test_command.sh
+@@ -0,0 +1,17 @@
++#!/bin/bash
++set -uo pipefail
++
++cd /testbed
++
++case "${1:-}" in
++  base)
++    go test github.com/azure/azure-dev/cli/azd/internal/vsrpc -v -count=1 -run '^(TestHandler|TestCancellation|TestObserverable)$'
++    ;;
++  new)
++    go test github.com/azure/azure-dev/cli/azd/internal/vsrpc -v -count=1 -run '^(TestPanicRecoveredAcrossHandlerAdapters|TestDebugServicePanicErrorDetails|TestDebugServicePanicDoesNotCloseConnection)$'
++    ;;
++  *)
++    echo "usage: $0 base|new"
++    exit 2
++    ;;
++esac

tasks/format-code-task-000034/tests/test.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+# Code grading, mimoagent opensource_code.py: reset the files the hidden test patch touches, apply it, run the
+# task's test command; reward 1 when it exits 0. A testbed failure (reset or apply) writes no reward: not scored.
+M=/var/lib/mimo
+CWD=/testbed
+V=/logs/verifier
+mkdir -p "$V"
+BASE=$(cat "$M/base" 2>/dev/null) || { echo "setup never ran (no $M/base)" >&2; exit 1; }
+cd "$CWD" || exit 1
+if [ -d "$M/git-hidden" ]; then rm -rf "$CWD/.git"; mv "$M/git-hidden" "$CWD/.git"; fi
+# the agent's change against the base, so its own commits count too
+git add -A >/dev/null 2>&1 && git -c core.fileMode=false diff --cached "$BASE" > "$V/agent.diff"
+git reset -q
+if ! (
+while IFS= read -r tf; do
+  [ -z "$tf" ] && continue
+  if git cat-file -e $BASE:"$tf" 2>/dev/null; then
+    git checkout $BASE -- "$tf" 2>/dev/null || true
+  else
+    git rm -f --cached "$tf" >/dev/null 2>&1 || true
+    rm -f "$tf"
+  fi
+done <<'EOF_RESET_TEST_FILES'
+cli/azd/internal/vsrpc/handler_test.go
+cli/azd/internal/vsrpc/server_test.go
+mimo_test_command.sh
+EOF_RESET_TEST_FILES
+
+); then echo "could not reset the test files (testbed problem, not scored)" >&2; exit 1; fi
+if ! git apply --verbose /tests/test.patch > "$V/apply.log" 2>&1; then
+  cat "$V/apply.log" >&2
+  echo "the hidden tests could not be applied (testbed problem, not scored)" >&2
+  exit 1
+fi
+timeout 1800 sh -c "$(cat /tests/test_command.sh)" > "$V/test_output.log" 2>&1
+RC=$?
+tail -c 12000 "$V/test_output.log"
+if [ $RC -eq 0 ]; then echo 1 > "$V/reward.txt"; else echo 0 > "$V/reward.txt"; fi
+echo "test command exited $RC"

tasks/format-code-task-000034/tests/test_command.sh

@@ -0,0 +1 @@
+bash /testbed/mimo_test_command.sh base && bash /testbed/mimo_test_command.sh new

tasks/format-code-task-000035/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f0d2850673fd9e046920cdde1d05976e8e941d2d4aa08931551ba0a76b7b8a40

tasks/format-code-task-000035/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com

tasks/format-code-task-000035/environment/setup/setup.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# Runs once, as root, before the agent starts: Harbor's [environment.healthcheck] unpacks this folder into
+# /var/lib/mimo and runs it. Generated by mimo_harbor; see its README for how each step maps to Xiaomi's harness.
+M=/var/lib/mimo
+[ -f "$M/ready" ] && exit 0
+chmod 700 "$M"
+fail() { echo "setup: $*" >&2; exit 1; }
+# the answer-leak blocklist (mimoagent), applied to /etc/hosts by the agent right after it installs
+write_blocklist() { cp "$M/files/blocklist" "$M/blocklist"; }
+
+# Code: mimoagent environments/datasets/opensource_code.py + base.py, as the explorer runs it (app/runner/domains.py Code).
+CWD=/testbed
+git config --global --add safe.directory "$CWD"
+cd "$CWD" || fail "no working directory $CWD"
+if ! git rev-parse --git-dir >/dev/null 2>&1; then   # a bare source tree: give it a baseline commit
+  git init -q && git add -A && git -c user.name=mimo -c user.email=mimo@localhost commit -q -m baseline --allow-empty
+fi
+BASE=$(git rev-parse HEAD 2>/dev/null)
+[ ${#BASE} -eq 40 ] || fail "could not resolve the base commit in $CWD"
+echo "$BASE" > "$M/base"
+# Images are built with history truncated at the base. If one is not, the fix could be read out of git log,
+# so .git is hidden while the agent works and put back for grading.
+LATER=$(git rev-list --all --not "$BASE" 2>/dev/null | head -n 5 | grep -c . || true)
+rm -f /tmp/fail.log /tmp/pass.log /tmp/patch.diff /tmp/test_patch.diff /tmp/*.log 2>/dev/null; rm -rf /tmp/claude-0 /tmp/claude-* /tmp/testem-* /tmp/puppeteer_dev_chrome_profile-* 2>/dev/null; rm -f /tmp/test_files.json 2>/dev/null; rm -rf /tmp/jest_* /tmp/jest-* /tmp/build /tmp/build_env 2>/dev/null; rm -rf /tmp/pytest-of-root /tmp/pytest-* /tmp/.pytest_cache /tmp/__pycache__ 2>/dev/null; rm -rf /tmp/go-build* /tmp/standards /tmp/wordpress /tmp/zig-* /tmp/testbase /tmp/testbed 2>/dev/null; rm -f /tmp/tmp*.tmp /tmp/*.bak /tmp/expect* /tmp/butwas* 2>/dev/null; rm -rf /root/.cache/go-build 2>/dev/null; rm -rf /tests 2>/dev/null; find /logs -mindepth 1 -delete 2>/dev/null; mkdir -p /logs/agent /logs/verifier /logs/artifacts; chmod 777 /logs/agent /logs/verifier /logs/artifacts; find /var/log -type f -delete 2>/dev/null || true; rm -rf /var/lib/postgresql/*/*/log /var/lib/postgresql/*/*/pg_log 2>/dev/null || true; find /var/lib/mysql /var/lib/mongodb -type f -name '*.log' -delete 2>/dev/null || true; rm -rf /root/.npm/_logs /root/.babel.json /root/.pytest_cache 2>/dev/null || true; find / -maxdepth 4 -xdev -name task_description.md -path '*/.build_env/*' -delete 2>/dev/null || true; true
+git clean -fdx --exclude=node_modules --exclude=bower_components --exclude=.husky --exclude=vendor --exclude=third_party --exclude=_deps --exclude=vcpkg_installed --exclude=.venv --exclude=venv --exclude=.gradle --exclude=target --exclude=.build --exclude=lib --exclude=.bundle --exclude=Manifest.toml --exclude=_build --exclude=.stack-work --exclude=dist-newstyle >/dev/null 2>&1
+find /root/.m2 -type f \( -name '*-SNAPSHOT.jar' -o -name '*-SNAPSHOT-sources.jar' -o -name '*-SNAPSHOT-tests.jar' -o -name '*-SNAPSHOT-test-sources.jar' \) -delete 2>/dev/null; rm -rf /root/.julia/compiled 2>/dev/null; rm -rf /root/.gradle/caches/build-cache-* /root/.gradle/daemon 2>/dev/null; rm -rf /root/.gradle/caches/*/scripts /root/.gradle/caches/jars-* 2>/dev/null; rm -rf /root/.cache/bazel 2>/dev/null; true
+if [ "$LATER" -gt 0 ]; then
+  mv "$CWD/.git" "$M/git-hidden"
+  echo "history not truncated at ${BASE:0:12}: .git hidden while the agent works"
+fi
+write_blocklist
+
+touch "$M/ready"
+echo "setup done"

tasks/format-code-task-000035/tests/test.patch

@@ -0,0 +1,200 @@
+diff --git a/packages/components/psammead-brand/src/print.test.jsx b/packages/components/psammead-brand/src/print.test.jsx
+new file mode 100644
+index 000000000..76d1d3445
+--- /dev/null
++++ b/packages/components/psammead-brand/src/print.test.jsx
+@@ -0,0 +1,172 @@
++import React from 'react';
++import renderer from 'react-test-renderer';
++import 'jest-styled-components';
++// eslint-disable-next-line import/no-extraneous-dependencies
++import { getCSS } from 'jest-styled-components/src/utils';
++import Brand from './index';
++
++const findByType = (node, type) => {
++  if (!node || typeof node !== 'object') return null;
++  if (Array.isArray(node)) {
++    for (let i = 0; i < node.length; i += 1) {
++      const f = findByType(node[i], type);
++      if (f) return f;
++    }
++    return null;
++  }
++  if (node.type === type) return node;
++  if (!node.children) return null;
++  for (let i = 0; i < node.children.length; i += 1) {
++    const f = findByType(node.children[i], type);
++    if (f) return f;
++  }
++  return null;
++};
++
++
++const getGClassTokens = tree => {
++  const g = findByType(tree, 'g');
++  if (!g || !g.props || !g.props.className) return [];
++  return String(g.props.className).trim().split(/\s+/);
++};
++
++
++const getWrapperClassTokens = tree => {
++  if (!tree || !tree.props || !tree.props.className) return [];
++  return String(tree.props.className).trim().split(/\s+/);
++};
++
++
++const collectPrintMediaRules = ast => {
++  if (!ast || !ast.stylesheet) return [];
++  return ast.stylesheet.rules.filter(
++    r => r.type === 'media' && /print/.test(r.media || ''),
++  );
++};
++
++
++
++const selectorTargetsG = (selector, gClassTokens) => {
++  if (typeof selector !== 'string') return false;
++  if (gClassTokens.some(c => selector.includes(`.${c}`))) return true;
++
++  if (/(^|[\s>+~])g\s*$/.test(selector.trim())) return true;
++  return false;
++};
++
++
++
++const selectorTargetsWrapper = (selector, wrapperClassTokens) =>
++  typeof selector === 'string' &&
++  wrapperClassTokens.some(c => selector.includes(`.${c}`));
++
++
++
++
++
++
++
++const hasValidPrintVisibilityFix = (ast, gClassTokens, wrapperClassTokens) => {
++  const printRules = collectPrintMediaRules(ast);
++  for (let i = 0; i < printRules.length; i += 1) {
++    const media = printRules[i];
++    if (!media.rules) continue; // eslint-disable-line no-continue
++    for (let j = 0; j < media.rules.length; j += 1) {
++      const rule = media.rules[j];
++      if (rule.type !== 'rule' || !rule.declarations) continue; // eslint-disable-line no-continue
++
++      const hasFill = rule.declarations.some(d => d.property === 'fill');
++      const hasBg = rule.declarations.some(
++        d => d.property === 'background-color',
++      );
++      const hasPrintColorAdjust = rule.declarations.some(
++        d =>
++          (d.property === 'print-color-adjust' ||
++            d.property === '-webkit-print-color-adjust') &&
++          d.value === 'exact',
++      );
++
++      const selectors = rule.selectors || [];
++
++
++      if (
++        hasFill &&
++        selectors.some(s => selectorTargetsG(s, gClassTokens))
++      ) {
++        return true;
++      }
++
++
++      if (
++        hasBg &&
++        hasPrintColorAdjust &&
++        selectors.some(s => selectorTargetsWrapper(s, wrapperClassTokens))
++      ) {
++        return true;
++      }
++    }
++  }
++  return false;
++};
++
++describe('Brand print visibility fix', () => {
++  it('ensures the SVG logo remains visible when printing', () => {
++    const tree = renderer.create(<Brand brandName="BBC News" />).toJSON();
++    const gClassTokens = getGClassTokens(tree);
++    const wrapperClassTokens = getWrapperClassTokens(tree);
++    const ast = getCSS();
++
++    expect(
++      hasValidPrintVisibilityFix(ast, gClassTokens, wrapperClassTokens),
++    ).toBe(true);
++  });
++
++  it('does not leave the <g> fill="#fff" attribute unmitigated', () => {
++
++
++
++
++
++    const tree = renderer.create(<Brand brandName="BBC News" />).toJSON();
++    const g = findByType(tree, 'g');
++    expect(g).not.toBeNull();
++
++    const fillAttr = g.props.fill;
++    const gClassTokens = getGClassTokens(tree);
++    const wrapperClassTokens = getWrapperClassTokens(tree);
++    const ast = getCSS();
++    const cssLevelFix = hasValidPrintVisibilityFix(
++      ast,
++      gClassTokens,
++      wrapperClassTokens,
++    );
++
++    const attrLevelFix = fillAttr !== '#fff' && fillAttr !== '#FFFFFF';
++    expect(cssLevelFix || attrLevelFix).toBe(true);
++  });
++
++  it('preserves the BBC News link href (regression check)', () => {
++    const tree = renderer
++      .create(<Brand brandName="BBC News" />)
++      .toJSON();
++    const a = findByType(tree, 'a');
++    expect(a).not.toBeNull();
++    expect(a.props.href).toBe('https://www.bbc.co.uk/news');
++  });
++
++  it('preserves the rendered brand name (regression check)', () => {
++    const tree = renderer
++      .create(<Brand brandName="My Branded Service" />)
++      .toJSON();
++    expect(JSON.stringify(tree)).toContain('My Branded Service');
++  });
++
++  it('preserves the SVG element with the expected viewBox (regression check)', () => {
++    const tree = renderer
++      .create(<Brand brandName="BBC News" />)
++      .toJSON();
++    const svg = findByType(tree, 'svg');
++    expect(svg).not.toBeNull();
++    expect(svg.props.viewBox).toBe('0 0 167.95 24');
++  });
++});
+diff --git a/test_commands.json b/test_commands.json
+new file mode 100644
+index 000000000..4393faeba
+--- /dev/null
++++ b/test_commands.json
+@@ -0,0 +1,3 @@
++{
++  "test_commands": ["npx jest packages/components/psammead-brand/src/print.test.jsx --verbose"]
++}
+diff --git a/mimo_test_command.sh b/mimo_test_command.sh
+new file mode 100755
+--- /dev/null
++++ b/mimo_test_command.sh
+@@ -0,0 +1,8 @@
++#!/usr/bin/env bash
++
++set -uo pipefail
++cd /testbed
++rc=0
++npx jest packages/components/psammead-brand/src/print.test.jsx --verbose
++rc=$(( rc | $? ))
++exit $rc

tasks/format-code-task-000035/tests/test.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+# Code grading, mimoagent opensource_code.py: reset the files the hidden test patch touches, apply it, run the
+# task's test command; reward 1 when it exits 0. A testbed failure (reset or apply) writes no reward: not scored.
+M=/var/lib/mimo
+CWD=/testbed
+V=/logs/verifier
+mkdir -p "$V"
+BASE=$(cat "$M/base" 2>/dev/null) || { echo "setup never ran (no $M/base)" >&2; exit 1; }
+cd "$CWD" || exit 1
+if [ -d "$M/git-hidden" ]; then rm -rf "$CWD/.git"; mv "$M/git-hidden" "$CWD/.git"; fi
+# the agent's change against the base, so its own commits count too
+git add -A >/dev/null 2>&1 && git -c core.fileMode=false diff --cached "$BASE" > "$V/agent.diff"
+git reset -q
+if ! (
+while IFS= read -r tf; do
+  [ -z "$tf" ] && continue
+  if git cat-file -e $BASE:"$tf" 2>/dev/null; then
+    git checkout $BASE -- "$tf" 2>/dev/null || true
+  else
+    git rm -f --cached "$tf" >/dev/null 2>&1 || true
+    rm -f "$tf"
+  fi
+done <<'EOF_RESET_TEST_FILES'
+packages/components/psammead-brand/src/print.test.jsx
+test_commands.json
+mimo_test_command.sh
+EOF_RESET_TEST_FILES
+
+); then echo "could not reset the test files (testbed problem, not scored)" >&2; exit 1; fi
+if ! git apply --verbose /tests/test.patch > "$V/apply.log" 2>&1; then
+  cat "$V/apply.log" >&2
+  echo "the hidden tests could not be applied (testbed problem, not scored)" >&2
+  exit 1
+fi
+timeout 1800 sh -c "$(cat /tests/test_command.sh)" > "$V/test_output.log" 2>&1
+RC=$?
+tail -c 12000 "$V/test_output.log"
+if [ $RC -eq 0 ]; then echo 1 > "$V/reward.txt"; else echo 0 > "$V/reward.txt"; fi
+echo "test command exited $RC"

tasks/format-code-task-000035/tests/test_command.sh

@@ -0,0 +1 @@
+bash /testbed/mimo_test_command.sh

tasks/format-code-task-000046/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:ceeb9bea228a86c93e3a983e9b9e1ec7ff8d761feccc194c57d110cc75b4e519

tasks/format-code-task-000046/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com

tasks/format-code-task-000046/environment/setup/setup.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# Runs once, as root, before the agent starts: Harbor's [environment.healthcheck] unpacks this folder into
+# /var/lib/mimo and runs it. Generated by mimo_harbor; see its README for how each step maps to Xiaomi's harness.
+M=/var/lib/mimo
+[ -f "$M/ready" ] && exit 0
+chmod 700 "$M"
+fail() { echo "setup: $*" >&2; exit 1; }
+# the answer-leak blocklist (mimoagent), applied to /etc/hosts by the agent right after it installs
+write_blocklist() { cp "$M/files/blocklist" "$M/blocklist"; }
+
+# Code: mimoagent environments/datasets/opensource_code.py + base.py, as the explorer runs it (app/runner/domains.py Code).
+CWD=/testbed
+git config --global --add safe.directory "$CWD"
+cd "$CWD" || fail "no working directory $CWD"
+if ! git rev-parse --git-dir >/dev/null 2>&1; then   # a bare source tree: give it a baseline commit
+  git init -q && git add -A && git -c user.name=mimo -c user.email=mimo@localhost commit -q -m baseline --allow-empty
+fi
+BASE=$(git rev-parse HEAD 2>/dev/null)
+[ ${#BASE} -eq 40 ] || fail "could not resolve the base commit in $CWD"
+echo "$BASE" > "$M/base"
+# Images are built with history truncated at the base. If one is not, the fix could be read out of git log,
+# so .git is hidden while the agent works and put back for grading.
+LATER=$(git rev-list --all --not "$BASE" 2>/dev/null | head -n 5 | grep -c . || true)
+rm -f /tmp/fail.log /tmp/pass.log /tmp/patch.diff /tmp/test_patch.diff /tmp/*.log 2>/dev/null; rm -rf /tmp/claude-0 /tmp/claude-* /tmp/testem-* /tmp/puppeteer_dev_chrome_profile-* 2>/dev/null; rm -f /tmp/test_files.json 2>/dev/null; rm -rf /tmp/jest_* /tmp/jest-* /tmp/build /tmp/build_env 2>/dev/null; rm -rf /tmp/pytest-of-root /tmp/pytest-* /tmp/.pytest_cache /tmp/__pycache__ 2>/dev/null; rm -rf /tmp/go-build* /tmp/standards /tmp/wordpress /tmp/zig-* /tmp/testbase /tmp/testbed 2>/dev/null; rm -f /tmp/tmp*.tmp /tmp/*.bak /tmp/expect* /tmp/butwas* 2>/dev/null; rm -rf /root/.cache/go-build 2>/dev/null; rm -rf /tests 2>/dev/null; find /logs -mindepth 1 -delete 2>/dev/null; mkdir -p /logs/agent /logs/verifier /logs/artifacts; chmod 777 /logs/agent /logs/verifier /logs/artifacts; find /var/log -type f -delete 2>/dev/null || true; rm -rf /var/lib/postgresql/*/*/log /var/lib/postgresql/*/*/pg_log 2>/dev/null || true; find /var/lib/mysql /var/lib/mongodb -type f -name '*.log' -delete 2>/dev/null || true; rm -rf /root/.npm/_logs /root/.babel.json /root/.pytest_cache 2>/dev/null || true; find / -maxdepth 4 -xdev -name task_description.md -path '*/.build_env/*' -delete 2>/dev/null || true; true
+git clean -fdx --exclude=node_modules --exclude=bower_components --exclude=.husky --exclude=vendor --exclude=third_party --exclude=_deps --exclude=vcpkg_installed --exclude=.venv --exclude=venv --exclude=.gradle --exclude=target --exclude=.build --exclude=lib --exclude=.bundle --exclude=Manifest.toml --exclude=_build --exclude=.stack-work --exclude=dist-newstyle >/dev/null 2>&1
+find /root/.m2 -type f \( -name '*-SNAPSHOT.jar' -o -name '*-SNAPSHOT-sources.jar' -o -name '*-SNAPSHOT-tests.jar' -o -name '*-SNAPSHOT-test-sources.jar' \) -delete 2>/dev/null; rm -rf /root/.julia/compiled 2>/dev/null; rm -rf /root/.gradle/caches/build-cache-* /root/.gradle/daemon 2>/dev/null; rm -rf /root/.gradle/caches/*/scripts /root/.gradle/caches/jars-* 2>/dev/null; rm -rf /root/.cache/bazel 2>/dev/null; true
+if [ "$LATER" -gt 0 ]; then
+  mv "$CWD/.git" "$M/git-hidden"
+  echo "history not truncated at ${BASE:0:12}: .git hidden while the agent works"
+fi
+write_blocklist
+
+touch "$M/ready"
+echo "setup done"

tasks/format-code-task-000046/tests/test.patch

@@ -0,0 +1,180 @@
+diff --git a/test/parse/grammar/interperter_spec.ts b/test/parse/grammar/interperter_spec.ts
+index 2d0629124..a7b0c0b7d 100644
+--- a/test/parse/grammar/interperter_spec.ts
++++ b/test/parse/grammar/interperter_spec.ts
+@@ -1,38 +1,39 @@
+ import {ITokenGrammarPath} from "../../../src/parse/grammar/path_public"
+ import {
+-    ActionTok,
+     actionDec,
++    actionDecSep,
++    ActionTok,
++    atLeastOneRule,
++    atLeastOneSepRule,
++    callArguments,
++    ColonTok,
++    CommaTok,
++    DotTok,
+     IdentTok,
+     LParenTok,
+-    RParenTok,
+-    ColonTok,
+-    SemicolonTok,
+     LSquareTok,
+-    RSquareTok,
+-    DotTok,
+-    CommaTok,
+     paramSpec,
+-    callArguments,
+-    actionDecSep,
+-    atLeastOneRule,
+     qualifiedNameSep,
+-    atLeastOneSepRule
++    RParenTok,
++    RSquareTok,
++    SemicolonTok
+ } from "./samples"
+ import {
+     NextAfterTokenWalker,
++    nextPossibleTokensAfter,
++    NextTerminalAfterAtLeastOneSepWalker,
+     NextTerminalAfterAtLeastOneWalker,
+-    NextTerminalAfterManyWalker,
+     NextTerminalAfterManySepWalker,
+-    NextTerminalAfterAtLeastOneSepWalker,
+-    possiblePathsFrom,
+-    nextPossibleTokensAfter
++    NextTerminalAfterManyWalker,
++    possiblePathsFrom
+ } from "../../../src/parse/grammar/interpreter"
+-import {setEquality, createRegularToken} from "../../utils/matchers"
++import {createRegularToken, setEquality} from "../../utils/matchers"
+ import {gast} from "../../../src/parse/grammar/gast_public"
+-import {Token, IToken} from "../../../src/scan/tokens_public"
++import {createToken, IToken, Token} from "../../../src/scan/tokens_public"
+ import {map} from "../../../src/utils/utils"
+-import {TokenConstructor} from "../../../src/scan/lexer_public"
++import {Lexer, TokenConstructor} from "../../../src/scan/lexer_public"
+ import {augmentTokenClasses, tokenStructuredMatcher} from "../../../src/scan/tokens"
++import {Parser} from "../../../src/parse/parser_public"
+ 
+ let RepetitionMandatory = gast.RepetitionMandatory
+ let Terminal = gast.Terminal
+@@ -571,11 +572,11 @@ describe("The chevrotain grammar interpreter capabilities", () => {
+             ], Comma),
+                 new gast.Terminal(Gamma)
+             ]
+-
+             expect(extractPartialPaths(possiblePathsFrom(rep, 1))).to.deep.equal([[Alpha], [Gamma]])
+             expect(extractPartialPaths(possiblePathsFrom(rep, 2))).to.deep.equal([[Alpha, Alpha], [Gamma]])
+-            expect(extractPartialPaths(possiblePathsFrom(rep, 3))).to.deep.equal([[Alpha, Alpha, Gamma], [Gamma]])
+-            expect(extractPartialPaths(possiblePathsFrom(rep, 4))).to.deep.equal([[Alpha, Alpha, Gamma], [Gamma]])
++            expect(extractPartialPaths(possiblePathsFrom(rep, 3))).to.deep.equal([[Alpha, Alpha, Comma], [Alpha, Alpha, Gamma], [Gamma]])
++            expect(extractPartialPaths(possiblePathsFrom(rep, 4))).to.deep.equal(
++                [[Alpha, Alpha, Comma, Alpha], [Alpha, Alpha, Gamma], [Gamma]])
+         })
+ 
+         it("Mandatory Repetition with Separator", () => {
+@@ -590,8 +591,9 @@ describe("The chevrotain grammar interpreter capabilities", () => {
+ 
+             expect(extractPartialPaths(possiblePathsFrom(repMandSep, 1))).to.deep.equal([[Alpha]])
+             expect(extractPartialPaths(possiblePathsFrom(repMandSep, 2))).to.deep.equal([[Alpha, Alpha]])
+-            expect(extractPartialPaths(possiblePathsFrom(repMandSep, 3))).to.deep.equal([[Alpha, Alpha, Gamma]])
+-            expect(extractPartialPaths(possiblePathsFrom(repMandSep, 4))).to.deep.equal([[Alpha, Alpha, Gamma]])
++            expect(extractPartialPaths(possiblePathsFrom(repMandSep, 3))).to.deep.equal([[Alpha, Alpha, Comma], [Alpha, Alpha, Gamma]])
++            expect(extractPartialPaths(possiblePathsFrom(repMandSep, 4))).to.deep.equal(
++                [[Alpha, Alpha, Comma, Alpha], [Alpha, Alpha, Gamma]])
+         })
+ 
+         it("NonTerminal", () => {
+@@ -882,5 +884,68 @@ describe("The chevrotain grammar interpreter capabilities", () => {
+             expect(nextPossibleTokensAfter(seq, INPUT([Alpha, Beta, Gamma]), tokenStructuredMatcher, 5)).to.be.empty
+         })
+     })
++})
++
++describe("issue 391 - WITH_SEP variants do not take SEP into account in lookahead", () => {
++    it("Reproduce issue", () => {
++
++        const LParen = createToken({name: "LParen", pattern: /\(/})
++        const RParen = createToken({name: "RParen", pattern: /\)/})
++        const Comma = createToken({name: "Comma", pattern: /,/})
++        const FatArrow = createToken({name: "FatArrow", pattern: /=>/})
++        const Identifier = createToken({name: "Identifier", pattern: /[a-zA-Z]+/})
++        const WhiteSpace = createToken({name: "WhiteSpace", pattern: /\s+/, group: Lexer.SKIPPED})
++
++        const allTokens = [WhiteSpace, LParen, RParen, Comma, FatArrow, Identifier]
++        const issue391Lexer = new Lexer(allTokens)
+ 
++        class Issue391Parser extends Parser {
++            constructor(input:Token[] = []) {
++                super(input, allTokens, {maxLookahead: 4});
++                (Parser as any).performSelfAnalysis(this)
++            }
++
++            topRule = this.RULE("topRule", () => {
++                return this.OR1([
++                    {
++
++                        ALT: () => {
++                            this.CONSUME1(LParen)
++                            this.MANY_SEP({
++                                SEP: Comma,
++                                DEF: () => {
++                                    this.CONSUME1(Identifier)
++                                }
++                            })
++                            this.CONSUME1(RParen)
++                            this.CONSUME1(FatArrow)
++                        }
++                    },
++                    {
++
++                        ALT: () => {
++                            this.CONSUME2(LParen)
++                            this.CONSUME2(Identifier)
++                            this.CONSUME2(RParen)
++                        }
++                    },
++                ])
++            })
++        }
++
++        expect(() => new Issue391Parser([])).to.not.throw("Ambiguous alternatives: <1 ,2>")
++        const myParser = new Issue391Parser([])
++
++        function testInput(input) {
++            const tokens = issue391Lexer.tokenize(input).tokens
++            myParser.input = tokens
++            myParser.topRule()
++            expect(myParser.errors).to.be.empty
++        }
++
++        testInput("(x, y) => ")
++        testInput("() =>")
++        testInput("(x) =>")
++        testInput("(x)")
++    })
+ })
+diff --git a/mimo_build_env.tar.gz.b64 b/mimo_build_env.tar.gz.b64
+new file mode 100644
+--- /dev/null
++++ b/mimo_build_env.tar.gz.b64
+@@ -0,0 +1,6 @@
++H4sIAAAAAAAAA+3SQU+DMBgGYM78ii/xbKEM2A/Y3YPzTqB0o4ZR0pbp/r0fmOhiortsMcb3OVBI
++268tb0Uzmb6t9HBMoltJ2boolpZ9bZd3WWQyW0m5zmWUSlmsy4iKm+3ozORD7YgiZ234adyl/j9K
++fOYftA+VsodDPbTCd9db41L+q1J+5F+WOeefl5LzT6+3he/98/xVS0vwjW7j+I62LyaojoKlB9tq
++8ewpS2lnHfG9GOtgGtObcCIe1dHTadRb5cwYKBMrkcetPt779wIDz+apc8kNzzS9Ph/u7eSU9sQX
++jebFfTyMrxS8msc/TgOFThPHstdh6acdF4jnmtXBtlOvfSIaMyT8pbqaRNKbZjlFMtbO62Tvar7F
++LjFD0G7Ujp+VH7Xi88S//cMBAAAAAAAAAAAAAAAAAAAAAK7oDZKdjQMAKAAA
+diff --git a/mimo_test_command.sh b/mimo_test_command.sh
+new file mode 100755
+--- /dev/null
++++ b/mimo_test_command.sh
+@@ -0,0 +1,6 @@
++#!/usr/bin/env bash
++
++set -uo pipefail
++cd /testbed || exit 1
++base64 -d mimo_build_env.tar.gz.b64 | tar -xzf - -C /testbed || exit 1
++exec bash -lc 'cd /testbed && bash -eo pipefail /testbed/.build_env/test_command.sh'

tasks/format-code-task-000046/tests/test.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+# Code grading, mimoagent opensource_code.py: reset the files the hidden test patch touches, apply it, run the
+# task's test command; reward 1 when it exits 0. A testbed failure (reset or apply) writes no reward: not scored.
+M=/var/lib/mimo
+CWD=/testbed
+V=/logs/verifier
+mkdir -p "$V"
+BASE=$(cat "$M/base" 2>/dev/null) || { echo "setup never ran (no $M/base)" >&2; exit 1; }
+cd "$CWD" || exit 1
+if [ -d "$M/git-hidden" ]; then rm -rf "$CWD/.git"; mv "$M/git-hidden" "$CWD/.git"; fi
+# the agent's change against the base, so its own commits count too
+git add -A >/dev/null 2>&1 && git -c core.fileMode=false diff --cached "$BASE" > "$V/agent.diff"
+git reset -q
+if ! (
+while IFS= read -r tf; do
+  [ -z "$tf" ] && continue
+  if git cat-file -e $BASE:"$tf" 2>/dev/null; then
+    git checkout $BASE -- "$tf" 2>/dev/null || true
+  else
+    git rm -f --cached "$tf" >/dev/null 2>&1 || true
+    rm -f "$tf"
+  fi
+done <<'EOF_RESET_TEST_FILES'
+test/parse/grammar/interperter_spec.ts
+mimo_build_env.tar.gz.b64
+mimo_test_command.sh
+EOF_RESET_TEST_FILES
+
+); then echo "could not reset the test files (testbed problem, not scored)" >&2; exit 1; fi
+if ! git apply --verbose /tests/test.patch > "$V/apply.log" 2>&1; then
+  cat "$V/apply.log" >&2
+  echo "the hidden tests could not be applied (testbed problem, not scored)" >&2
+  exit 1
+fi
+timeout 1800 sh -c "$(cat /tests/test_command.sh)" > "$V/test_output.log" 2>&1
+RC=$?
+tail -c 12000 "$V/test_output.log"
+if [ $RC -eq 0 ]; then echo 1 > "$V/reward.txt"; else echo 0 > "$V/reward.txt"; fi
+echo "test command exited $RC"

tasks/format-code-task-000046/tests/test_command.sh

@@ -0,0 +1 @@
+bash /testbed/mimo_test_command.sh

tasks/format-code-task-000051/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:f190d89c473987498ee878e62c22a35f63f1a081bc5045e031430d2cdb13d0dc

tasks/format-code-task-000051/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com

tasks/format-code-task-000051/environment/setup/setup.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# Runs once, as root, before the agent starts: Harbor's [environment.healthcheck] unpacks this folder into
+# /var/lib/mimo and runs it. Generated by mimo_harbor; see its README for how each step maps to Xiaomi's harness.
+M=/var/lib/mimo
+[ -f "$M/ready" ] && exit 0
+chmod 700 "$M"
+fail() { echo "setup: $*" >&2; exit 1; }
+# the answer-leak blocklist (mimoagent), applied to /etc/hosts by the agent right after it installs
+write_blocklist() { cp "$M/files/blocklist" "$M/blocklist"; }
+
+# Code: mimoagent environments/datasets/opensource_code.py + base.py, as the explorer runs it (app/runner/domains.py Code).
+CWD=/testbed
+git config --global --add safe.directory "$CWD"
+cd "$CWD" || fail "no working directory $CWD"
+if ! git rev-parse --git-dir >/dev/null 2>&1; then   # a bare source tree: give it a baseline commit
+  git init -q && git add -A && git -c user.name=mimo -c user.email=mimo@localhost commit -q -m baseline --allow-empty
+fi
+BASE=$(git rev-parse HEAD 2>/dev/null)
+[ ${#BASE} -eq 40 ] || fail "could not resolve the base commit in $CWD"
+echo "$BASE" > "$M/base"
+# Images are built with history truncated at the base. If one is not, the fix could be read out of git log,
+# so .git is hidden while the agent works and put back for grading.
+LATER=$(git rev-list --all --not "$BASE" 2>/dev/null | head -n 5 | grep -c . || true)
+rm -f /tmp/fail.log /tmp/pass.log /tmp/patch.diff /tmp/test_patch.diff /tmp/*.log 2>/dev/null; rm -rf /tmp/claude-0 /tmp/claude-* /tmp/testem-* /tmp/puppeteer_dev_chrome_profile-* 2>/dev/null; rm -f /tmp/test_files.json 2>/dev/null; rm -rf /tmp/jest_* /tmp/jest-* /tmp/build /tmp/build_env 2>/dev/null; rm -rf /tmp/pytest-of-root /tmp/pytest-* /tmp/.pytest_cache /tmp/__pycache__ 2>/dev/null; rm -rf /tmp/go-build* /tmp/standards /tmp/wordpress /tmp/zig-* /tmp/testbase /tmp/testbed 2>/dev/null; rm -f /tmp/tmp*.tmp /tmp/*.bak /tmp/expect* /tmp/butwas* 2>/dev/null; rm -rf /root/.cache/go-build 2>/dev/null; rm -rf /tests 2>/dev/null; find /logs -mindepth 1 -delete 2>/dev/null; mkdir -p /logs/agent /logs/verifier /logs/artifacts; chmod 777 /logs/agent /logs/verifier /logs/artifacts; find /var/log -type f -delete 2>/dev/null || true; rm -rf /var/lib/postgresql/*/*/log /var/lib/postgresql/*/*/pg_log 2>/dev/null || true; find /var/lib/mysql /var/lib/mongodb -type f -name '*.log' -delete 2>/dev/null || true; rm -rf /root/.npm/_logs /root/.babel.json /root/.pytest_cache 2>/dev/null || true; find / -maxdepth 4 -xdev -name task_description.md -path '*/.build_env/*' -delete 2>/dev/null || true; true
+git clean -fdx --exclude=node_modules --exclude=bower_components --exclude=.husky --exclude=vendor --exclude=third_party --exclude=_deps --exclude=vcpkg_installed --exclude=.venv --exclude=venv --exclude=.gradle --exclude=target --exclude=.build --exclude=lib --exclude=.bundle --exclude=Manifest.toml --exclude=_build --exclude=.stack-work --exclude=dist-newstyle >/dev/null 2>&1
+find /root/.m2 -type f \( -name '*-SNAPSHOT.jar' -o -name '*-SNAPSHOT-sources.jar' -o -name '*-SNAPSHOT-tests.jar' -o -name '*-SNAPSHOT-test-sources.jar' \) -delete 2>/dev/null; rm -rf /root/.julia/compiled 2>/dev/null; rm -rf /root/.gradle/caches/build-cache-* /root/.gradle/daemon 2>/dev/null; rm -rf /root/.gradle/caches/*/scripts /root/.gradle/caches/jars-* 2>/dev/null; rm -rf /root/.cache/bazel 2>/dev/null; true
+if [ "$LATER" -gt 0 ]; then
+  mv "$CWD/.git" "$M/git-hidden"
+  echo "history not truncated at ${BASE:0:12}: .git hidden while the agent works"
+fi
+write_blocklist
+
+touch "$M/ready"
+echo "setup done"

tasks/format-code-task-000051/tests/test.patch

@@ -0,0 +1,499 @@
+diff --git a/linux_os/guide/system/permissions/files/permissions_var_log_dir/file_owner_var_log_syslog/tests/correct_owner.pass.sh b/linux_os/guide/system/permissions/files/permissions_var_log_dir/file_owner_var_log_syslog/tests/correct_owner.pass.sh
+new file mode 100644
+index 0000000000..5dbd79c4e6
+--- /dev/null
++++ b/linux_os/guide/system/permissions/files/permissions_var_log_dir/file_owner_var_log_syslog/tests/correct_owner.pass.sh
+@@ -0,0 +1,14 @@
++#!/bin/bash
++
++
++if [ ! -f /var/log/syslog ]; then
++    mkdir -p /var/log/
++    touch /var/log/syslog
++fi
++
++if ! grep -q "^syslog:" /etc/passwd; then
++    useradd syslog
++fi
++
++chown syslog /var/log/syslog
++
+diff --git a/linux_os/guide/system/permissions/files/permissions_var_log_dir/file_owner_var_log_syslog/tests/incorrect_owner.fail.sh b/linux_os/guide/system/permissions/files/permissions_var_log_dir/file_owner_var_log_syslog/tests/incorrect_owner.fail.sh
+new file mode 100644
+index 0000000000..c1855cbd56
+--- /dev/null
++++ b/linux_os/guide/system/permissions/files/permissions_var_log_dir/file_owner_var_log_syslog/tests/incorrect_owner.fail.sh
+@@ -0,0 +1,14 @@
++#!/bin/bash
++
++
++if [ ! -f /var/log/syslog ]; then
++    mkdir -p /var/log/
++    touch /var/log/syslog
++fi
++
++if ! grep -q "^syslog:" /etc/passwd; then
++    useradd syslog
++fi
++
++useradd testuser_123
++chown testuser_123 /var/log/syslog
+diff --git a/shared/templates/file_owner/tests/correct_owner.pass.sh b/shared/templates/file_owner/tests/correct_owner.pass.sh
+index f05418c94f..92b31b7d87 100644
+--- a/shared/templates/file_owner/tests/correct_owner.pass.sh
++++ b/shared/templates/file_owner/tests/correct_owner.pass.sh
+@@ -8,15 +8,15 @@ fi
+ {{% if FILE_REGEX %}}
+ echo "Create specific tests for this rule because of regex"
+ {{% elif RECURSIVE %}}
+-find -L {{{ path }}} -type d -exec chown {{{ FILEUID }}} {} \;
++find -L {{{ path }}} -type d -exec chown {{{ UID_OR_NAME }}} {} \;
+ {{% else %}}
+-chown {{{ FILEUID }}} {{{ path }}}
++chown {{{ UID_OR_NAME }}} {{{ path }}}
+ {{% endif %}}
+ {{% else %}}
+ if [ ! -f {{{ path }}} ]; then
+     mkdir -p "$(dirname '{{{ path }}}')"
+     touch {{{ path }}}
+ fi
+-chown {{{ FILEUID }}} {{{ path }}}
++chown {{{ UID_OR_NAME }}} {{{ path }}}
+ {{% endif %}}
+ {{% endfor %}}
+diff --git a/shared/templates/file_owner/tests/incorrect_owner.fail.sh b/shared/templates/file_owner/tests/incorrect_owner.fail.sh
+index 8d4ccfe0d5..82c16cb89a 100644
+--- a/shared/templates/file_owner/tests/incorrect_owner.fail.sh
++++ b/shared/templates/file_owner/tests/incorrect_owner.fail.sh
+@@ -13,6 +13,7 @@ useradd testuser_123
+ if [ ! -d {{{ path }}} ]; then
+     mkdir -p {{{ path }}}
+ fi
++touch "{{{ path }}}"/cac_file_owner_test_file
+ {{% if FILE_REGEX %}}
+ find -L {{{ path }}} {{{ FIND_RECURSE_ARGS }}} -type f -regex '{{{ FILE_REGEX[loop.index0] }}}' -exec chown testuser_123 {} \;
+ {{% elif RECURSIVE %}}
+diff --git a/tests/test_file_owner_uid_or_name_behavior.py b/tests/test_file_owner_uid_or_name_behavior.py
+new file mode 100644
+index 0000000000..6fa46dc554
+--- /dev/null
++++ b/tests/test_file_owner_uid_or_name_behavior.py
+@@ -0,0 +1,386 @@
++#!/usr/bin/env python3
++import importlib.util
++import os
++import random
++import re
++import shutil
++import subprocess
++import sys
++import tempfile
++import xml.etree.ElementTree as ET
++from pathlib import Path
++
++
++REPO_ROOT = Path(__file__).resolve().parents[1]
++OPENSHIFT_ROOT = REPO_ROOT / "applications" / "openshift"
++TEMPLATE_ROOT = REPO_ROOT / "shared" / "templates" / "file_owner"
++RULE_ID = "file_owner_fixture_uid_or_name"
++
++sys.path.insert(0, str(REPO_ROOT))
++
++import ssg.jinja  # noqa: E402
++
++
++def _assert(condition, message):
++    if not condition:
++        raise AssertionError(message)
++
++
++def _run(cmd, **kwargs):
++    result = subprocess.run(
++        cmd,
++        text=True,
++        stdout=subprocess.PIPE,
++        stderr=subprocess.PIPE,
++        **kwargs,
++    )
++    if result.returncode != 0:
++        raise AssertionError(
++            "command failed: {0}\nstdout:\n{1}\nstderr:\n{2}".format(
++                " ".join(cmd), result.stdout, result.stderr
++            )
++        )
++    return result
++
++
++def _load_template_preprocessor():
++
++
++
++    path = TEMPLATE_ROOT / "template.py"
++    spec = importlib.util.spec_from_file_location("fixture_file_owner_template", path)
++    module = importlib.util.module_from_spec(spec)
++    spec.loader.exec_module(module)
++    return module
++
++
++def _render_file_owner_template(lang, template_vars):
++    params = dict(template_vars)
++    params.setdefault("_rule_id", RULE_ID)
++    params.setdefault("recursive", "false")
++
++    preprocessor = _load_template_preprocessor()
++    processed = preprocessor.preprocess(params, lang)
++    substitutions = {key.upper(): value for key, value in processed.items()}
++    merged = ssg.jinja.load_macros()
++    merged.update(substitutions)
++
++    return ssg.jinja.process_file(str(TEMPLATE_ROOT / f"{lang}.template"), merged)
++
++
++def _parse_oval_fragment(rendered_oval):
++    wrapped = (
++        '<root xmlns:unix="urn:unix" xmlns:ind="urn:ind" xmlns:xsi="urn:xsi">'
++        + rendered_oval
++        + "</root>"
++    )
++    return ET.fromstring(wrapped)
++
++
++def _namespace(tag):
++    return {
++        "unix": f"{{urn:unix}}{tag}",
++        "ind": f"{{urn:ind}}{tag}",
++    }
++
++
++def _local_name(tag):
++    return tag.rsplit("}", 1)[-1]
++
++
++def _unique_user():
++    return "cacfo{0}{1}".format(os.getpid(), random.randrange(1000, 9999))
++
++
++def _ensure_user(name):
++    try:
++        import pwd
++
++        return pwd.getpwnam(name).pw_uid
++    except KeyError:
++        _run(["useradd", name])
++        return pwd.getpwnam(name).pw_uid
++
++
++def _delete_user(name):
++    subprocess.run(["userdel", "-r", name], text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
++
++
++def _run_shell_script(script):
++    result = subprocess.run(
++        ["bash", "-e"],
++        input=script,
++        text=True,
++        stdout=subprocess.PIPE,
++        stderr=subprocess.PIPE,
++    )
++    if result.returncode != 0:
++        raise AssertionError(
++            "rendered shell remediation failed\nstdout:\n{0}\nstderr:\n{1}\nscript:\n{2}".format(
++                result.stdout, result.stderr, script
++            )
++        )
++
++
++def test_bash_remediation_accepts_numeric_uid_and_username():
++    _assert(os.geteuid() == 0, "ownership behavior checks must run as root")
++    _assert(shutil.which("useradd"), "useradd is required for username ownership checks")
++    _assert(shutil.which("userdel"), "userdel is required for username ownership cleanup")
++
++    owner_name = _unique_user()
++    try:
++        owner_uid = _ensure_user(owner_name)
++        _assert(owner_uid != 0, "dynamic test account should not be root")
++
++        with tempfile.TemporaryDirectory(prefix="cac-file-owner-") as tmpdir:
++            target = Path(tmpdir) / "owned_file"
++            target.write_text("fixture file_owner behavior\n", encoding="utf-8")
++
++            os.chown(target, 0, -1)
++            numeric_script = _render_file_owner_template(
++                "bash", {"filepath": str(target), "uid_or_name": str(owner_uid)}
++            )
++            _run_shell_script(numeric_script)
++            _assert(
++                target.stat().st_uid == owner_uid,
++                "numeric uid_or_name should be accepted by the generated remediation",
++            )
++
++            os.chown(target, 0, -1)
++            username_script = _render_file_owner_template(
++                "bash", {"filepath": str(target), "uid_or_name": owner_name}
++            )
++            _run_shell_script(username_script)
++            _assert(
++                target.stat().st_uid == owner_uid,
++                "username uid_or_name should be resolved by the system when remediation runs",
++            )
++    finally:
++        _delete_user(owner_name)
++
++
++def test_oval_check_numeric_uid_uses_integer_user_id():
++    for uid in ("0", "4242"):
++        rendered = _render_file_owner_template(
++            "oval", {"filepath": "/tmp/cac-file-owner-fixture", "uid_or_name": uid}
++        )
++        root = _parse_oval_fragment(rendered)
++        user_ids = list(root.iter(_namespace("user_id")["unix"]))
++        _assert(user_ids, "rendered numeric OVAL must contain a file owner user_id state")
++        _assert(
++            any(el.attrib.get("datatype") == "int" and (el.text or "").strip() == uid for el in user_ids),
++            "numeric uid_or_name should render as an integer user_id matching the configured UID",
++        )
++        _assert(
++            all("var_ref" not in el.attrib for el in user_ids),
++            "numeric uid_or_name should not require a runtime username lookup variable",
++        )
++        passwd_lookups = [
++            (el.text or "").strip()
++            for el in root.iter(_namespace("filepath")["ind"])
++            if (el.text or "").strip() == "/etc/passwd"
++        ]
++        _assert(
++            not passwd_lookups,
++            "numeric uid_or_name should keep the previous direct numeric UID behavior",
++        )
++
++
++def test_oval_check_username_resolves_owner_on_scanned_system():
++    username = "cac_fixture_owner_{0}".format(random.randrange(10000, 99999))
++    rendered = _render_file_owner_template(
++        "oval", {"filepath": "/tmp/cac-file-owner-fixture", "uid_or_name": username}
++    )
++    root = _parse_oval_fragment(rendered)
++
++    user_ids = list(root.iter(_namespace("user_id")["unix"]))
++    _assert(user_ids, "rendered username OVAL must contain a file owner user_id state")
++    _assert(
++        any("var_ref" in el.attrib for el in user_ids),
++        "username uid_or_name should compare file owner against a runtime-resolved UID",
++    )
++    _assert(
++        all((el.text or "").strip() != username for el in user_ids),
++        "username uid_or_name must not be emitted as a literal integer user_id",
++    )
++
++    var_refs = {
++        el.attrib["var_ref"]
++        for el in user_ids
++        if el.attrib.get("datatype") == "int" and "var_ref" in el.attrib
++    }
++    _assert(
++        var_refs,
++        "username uid_or_name should use a runtime variable for the expected UID",
++    )
++
++    variables = {el.attrib.get("id"): el for el in root.iter() if _local_name(el.tag).endswith("variable")}
++    connected_variables = [variables[var_ref] for var_ref in var_refs if var_ref in variables]
++    _assert(
++        connected_variables,
++        "username UID comparison should be backed by an OVAL variable definition",
++    )
++
++    component_refs = set()
++    for variable in connected_variables:
++        for element in variable.iter():
++            ref = element.attrib.get("object_ref")
++            if ref:
++                component_refs.add(ref)
++
++    lookup_objects = {el.attrib.get("id"): el for el in root.iter(_namespace("textfilecontent54_object")["ind"])}
++    connected_lookup_objects = [lookup_objects[ref] for ref in component_refs if ref in lookup_objects]
++    _assert(
++        connected_lookup_objects,
++        "runtime UID variable should be derived from scanned system account data",
++    )
++
++    lookup_text = "\n".join(
++        (element.text or "").strip()
++        for lookup_object in connected_lookup_objects
++        for element in lookup_object.iter()
++    )
++    _assert(
++        username in lookup_text,
++        "runtime UID lookup should be tied to the configured account name",
++    )
++
++
++def _line_indent(line):
++    return len(line) - len(line.lstrip(" "))
++
++
++def _strip_quotes(value):
++    value = value.strip()
++    if len(value) >= 2 and value[0] in ("'", '"') and value[-1] == value[0]:
++        return value[1:-1]
++    return value
++
++
++def _active_file_owner_template_vars(rule_path):
++    lines = rule_path.read_text(encoding="utf-8").splitlines()
++    for index, line in enumerate(lines):
++        stripped = line.strip()
++        if not stripped or stripped.startswith("#"):
++            continue
++        if stripped != "template:":
++            continue
++
++        template_indent = _line_indent(line)
++        block = []
++        for candidate in lines[index + 1 :]:
++            candidate_stripped = candidate.strip()
++            if (
++                candidate_stripped
++                and not candidate_stripped.startswith("#")
++                and _line_indent(candidate) <= template_indent
++            ):
++                break
++            block.append(candidate)
++
++        is_file_owner = False
++        vars_index = None
++        vars_indent = None
++        for block_index, block_line in enumerate(block):
++            block_stripped = block_line.strip()
++            if not block_stripped or block_stripped.startswith("#") or block_stripped.startswith("{{"):
++                continue
++            if block_stripped.startswith("name:"):
++                is_file_owner = _strip_quotes(block_stripped.split(":", 1)[1]) == "file_owner"
++            if block_stripped == "vars:":
++                vars_index = block_index
++                vars_indent = _line_indent(block_line)
++
++        if not is_file_owner or vars_index is None:
++            continue
++
++        vars_ = {}
++        for var_line in block[vars_index + 1 :]:
++            var_stripped = var_line.strip()
++            if not var_stripped or var_stripped.startswith("#") or var_stripped.startswith("{{"):
++                continue
++            if _line_indent(var_line) <= vars_indent:
++                break
++            match = re.match(r"([A-Za-z0-9_@-]+)\s*:\s*(.*)$", var_stripped)
++            if match:
++                vars_[match.group(1)] = _strip_quotes(match.group(2))
++        return vars_
++    return None
++
++
++def _base_file_owner_template_vars(relative_path):
++    try:
++        result = _run(["git", "show", "HEAD:{0}".format(relative_path)])
++    except AssertionError:
++        return None
++
++    with tempfile.NamedTemporaryFile("w", encoding="utf-8", delete=False) as tmp:
++        tmp.write(result.stdout)
++        tmp_path = Path(tmp.name)
++    try:
++        return _active_file_owner_template_vars(tmp_path)
++    finally:
++        tmp_path.unlink(missing_ok=True)
++
++
++def test_openshift_file_owner_rules_use_uid_or_name_and_preserve_values():
++    discovered = {}
++    migrated_from_base = {}
++    offenders = []
++    for rule_path in OPENSHIFT_ROOT.rglob("rule.yml"):
++        vars_ = _active_file_owner_template_vars(rule_path)
++        if vars_ is None:
++            continue
++        relative = rule_path.relative_to(REPO_ROOT).as_posix()
++        discovered[relative] = vars_
++        base_vars = _base_file_owner_template_vars(relative)
++        if base_vars and "fileuid" in base_vars:
++            migrated_from_base[relative] = base_vars["fileuid"]
++        if "fileuid" in vars_ or "uid_or_name" not in vars_:
++            offenders.append(
++                "{0}: fileuid={1!r}, uid_or_name={2!r}".format(
++                    relative, vars_.get("fileuid"), vars_.get("uid_or_name")
++                )
++            )
++
++    _assert(discovered, "expected active OpenShift file_owner rules to exist")
++    _assert(
++        not offenders,
++        "active OpenShift file_owner rules must use uid_or_name instead of fileuid:\n"
++        + "\n".join(offenders),
++    )
++
++    _assert(
++        migrated_from_base,
++        "expected base OpenShift file_owner rules using fileuid to be migration targets",
++    )
++
++    wrong = []
++    for relative, expected in sorted(migrated_from_base.items()):
++        if relative not in discovered:
++            wrong.append("{0}: active file_owner rule disappeared".format(relative))
++            continue
++        actual = str(discovered[relative].get("uid_or_name"))
++        if actual != expected:
++            wrong.append("{0}: expected {1!r}, got {2!r}".format(relative, expected, actual))
++    _assert(
++        not wrong,
++        "OpenShift file_owner uid_or_name values must preserve the previous owner:\n"
++        + "\n".join(wrong),
++    )
++
++
++def main():
++    tests = [
++        test_bash_remediation_accepts_numeric_uid_and_username,
++        test_oval_check_numeric_uid_uses_integer_user_id,
++        test_oval_check_username_resolves_owner_on_scanned_system,
++        test_openshift_file_owner_rules_use_uid_or_name_and_preserve_values,
++    ]
++    for test in tests:
++        test()
++        print("PASS {0}".format(test.__name__))
++
++
++if __name__ == "__main__":
++    main()
+diff --git a/mimo_test_command.sh b/mimo_test_command.sh
+new file mode 100755
+--- /dev/null
++++ b/mimo_test_command.sh
+@@ -0,0 +1,27 @@
++#!/bin/bash
++set -uo pipefail
++cd /testbed
++
++
++SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
++REPO_ROOT="/testbed"
++
++
++
++case "${1:-}" in
++  base)
++    python3 - <<'PY'
++import sys
++sys.path.insert(0, "/testbed")
++import ssg.jinja
++print("base smoke: ssg.jinja import ok")
++PY
++    ;;
++  new)
++    python3 /testbed/tests/test_file_owner_uid_or_name_behavior.py
++    ;;
++  *)
++    echo "usage: $0 base|new"
++    exit 2
++    ;;
++esac

tasks/format-code-task-000051/tests/test.sh

@@ -0,0 +1,42 @@
+#!/bin/bash
+# Code grading, mimoagent opensource_code.py: reset the files the hidden test patch touches, apply it, run the
+# task's test command; reward 1 when it exits 0. A testbed failure (reset or apply) writes no reward: not scored.
+M=/var/lib/mimo
+CWD=/testbed
+V=/logs/verifier
+mkdir -p "$V"
+BASE=$(cat "$M/base" 2>/dev/null) || { echo "setup never ran (no $M/base)" >&2; exit 1; }
+cd "$CWD" || exit 1
+if [ -d "$M/git-hidden" ]; then rm -rf "$CWD/.git"; mv "$M/git-hidden" "$CWD/.git"; fi
+# the agent's change against the base, so its own commits count too
+git add -A >/dev/null 2>&1 && git -c core.fileMode=false diff --cached "$BASE" > "$V/agent.diff"
+git reset -q
+if ! (
+while IFS= read -r tf; do
+  [ -z "$tf" ] && continue
+  if git cat-file -e $BASE:"$tf" 2>/dev/null; then
+    git checkout $BASE -- "$tf" 2>/dev/null || true
+  else
+    git rm -f --cached "$tf" >/dev/null 2>&1 || true
+    rm -f "$tf"
+  fi
+done <<'EOF_RESET_TEST_FILES'
+linux_os/guide/system/permissions/files/permissions_var_log_dir/file_owner_var_log_syslog/tests/correct_owner.pass.sh
+linux_os/guide/system/permissions/files/permissions_var_log_dir/file_owner_var_log_syslog/tests/incorrect_owner.fail.sh
+shared/templates/file_owner/tests/correct_owner.pass.sh
+shared/templates/file_owner/tests/incorrect_owner.fail.sh
+tests/test_file_owner_uid_or_name_behavior.py
+mimo_test_command.sh
+EOF_RESET_TEST_FILES
+
+); then echo "could not reset the test files (testbed problem, not scored)" >&2; exit 1; fi
+if ! git apply --verbose /tests/test.patch > "$V/apply.log" 2>&1; then
+  cat "$V/apply.log" >&2
+  echo "the hidden tests could not be applied (testbed problem, not scored)" >&2
+  exit 1
+fi
+timeout 1800 sh -c "$(cat /tests/test_command.sh)" > "$V/test_output.log" 2>&1
+RC=$?
+tail -c 12000 "$V/test_output.log"
+if [ $RC -eq 0 ]; then echo 1 > "$V/reward.txt"; else echo 0 > "$V/reward.txt"; fi
+echo "test command exited $RC"

tasks/format-code-task-000051/tests/test_command.sh

@@ -0,0 +1 @@
+bash /testbed/mimo_test_command.sh base && bash /testbed/mimo_test_command.sh new

tasks/format-code-task-000056/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:d026ee734aefb4a241fb86d03b1543ebd95f6a7259109c02a4be47500d8b2f74

tasks/format-code-task-000056/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com

tasks/format-code-task-000056/environment/setup/setup.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# Runs once, as root, before the agent starts: Harbor's [environment.healthcheck] unpacks this folder into
+# /var/lib/mimo and runs it. Generated by mimo_harbor; see its README for how each step maps to Xiaomi's harness.
+M=/var/lib/mimo
+[ -f "$M/ready" ] && exit 0
+chmod 700 "$M"
+fail() { echo "setup: $*" >&2; exit 1; }
+# the answer-leak blocklist (mimoagent), applied to /etc/hosts by the agent right after it installs
+write_blocklist() { cp "$M/files/blocklist" "$M/blocklist"; }
+
+# Code: mimoagent environments/datasets/opensource_code.py + base.py, as the explorer runs it (app/runner/domains.py Code).
+CWD=/testbed
+git config --global --add safe.directory "$CWD"
+cd "$CWD" || fail "no working directory $CWD"
+if ! git rev-parse --git-dir >/dev/null 2>&1; then   # a bare source tree: give it a baseline commit
+  git init -q && git add -A && git -c user.name=mimo -c user.email=mimo@localhost commit -q -m baseline --allow-empty
+fi
+BASE=$(git rev-parse HEAD 2>/dev/null)
+[ ${#BASE} -eq 40 ] || fail "could not resolve the base commit in $CWD"
+echo "$BASE" > "$M/base"
+# Images are built with history truncated at the base. If one is not, the fix could be read out of git log,
+# so .git is hidden while the agent works and put back for grading.
+LATER=$(git rev-list --all --not "$BASE" 2>/dev/null | head -n 5 | grep -c . || true)
+rm -f /tmp/fail.log /tmp/pass.log /tmp/patch.diff /tmp/test_patch.diff /tmp/*.log 2>/dev/null; rm -rf /tmp/claude-0 /tmp/claude-* /tmp/testem-* /tmp/puppeteer_dev_chrome_profile-* 2>/dev/null; rm -f /tmp/test_files.json 2>/dev/null; rm -rf /tmp/jest_* /tmp/jest-* /tmp/build /tmp/build_env 2>/dev/null; rm -rf /tmp/pytest-of-root /tmp/pytest-* /tmp/.pytest_cache /tmp/__pycache__ 2>/dev/null; rm -rf /tmp/go-build* /tmp/standards /tmp/wordpress /tmp/zig-* /tmp/testbase /tmp/testbed 2>/dev/null; rm -f /tmp/tmp*.tmp /tmp/*.bak /tmp/expect* /tmp/butwas* 2>/dev/null; rm -rf /root/.cache/go-build 2>/dev/null; rm -rf /tests 2>/dev/null; find /logs -mindepth 1 -delete 2>/dev/null; mkdir -p /logs/agent /logs/verifier /logs/artifacts; chmod 777 /logs/agent /logs/verifier /logs/artifacts; find /var/log -type f -delete 2>/dev/null || true; rm -rf /var/lib/postgresql/*/*/log /var/lib/postgresql/*/*/pg_log 2>/dev/null || true; find /var/lib/mysql /var/lib/mongodb -type f -name '*.log' -delete 2>/dev/null || true; rm -rf /root/.npm/_logs /root/.babel.json /root/.pytest_cache 2>/dev/null || true; find / -maxdepth 4 -xdev -name task_description.md -path '*/.build_env/*' -delete 2>/dev/null || true; true
+git clean -fdx --exclude=node_modules --exclude=bower_components --exclude=.husky --exclude=vendor --exclude=third_party --exclude=_deps --exclude=vcpkg_installed --exclude=.venv --exclude=venv --exclude=.gradle --exclude=target --exclude=.build --exclude=lib --exclude=.bundle --exclude=Manifest.toml --exclude=_build --exclude=.stack-work --exclude=dist-newstyle >/dev/null 2>&1
+find /root/.m2 -type f \( -name '*-SNAPSHOT.jar' -o -name '*-SNAPSHOT-sources.jar' -o -name '*-SNAPSHOT-tests.jar' -o -name '*-SNAPSHOT-test-sources.jar' \) -delete 2>/dev/null; rm -rf /root/.julia/compiled 2>/dev/null; rm -rf /root/.gradle/caches/build-cache-* /root/.gradle/daemon 2>/dev/null; rm -rf /root/.gradle/caches/*/scripts /root/.gradle/caches/jars-* 2>/dev/null; rm -rf /root/.cache/bazel 2>/dev/null; true
+if [ "$LATER" -gt 0 ]; then
+  mv "$CWD/.git" "$M/git-hidden"
+  echo "history not truncated at ${BASE:0:12}: .git hidden while the agent works"
+fi
+write_blocklist
+
+touch "$M/ready"
+echo "setup done"

tasks/format-code-task-000056/tests/test.patch

@@ -0,0 +1,288 @@
+diff --git a/pkg/serverless/logs/logs_test.go b/pkg/serverless/logs/logs_test.go
+index 938393f354..19a88406ae 100644
+--- a/pkg/serverless/logs/logs_test.go
++++ b/pkg/serverless/logs/logs_test.go
+@@ -12,6 +12,7 @@ import (
+ 	"net/http"
+ 	"net/http/httptest"
+ 	"os"
++	"strings"
+ 	"testing"
+ 	"time"
+ 
+@@ -392,7 +393,74 @@ func TestProcessMessageShouldNotProcessLogsDropped(t *testing.T) {
+ 	assert.Len(t, timed, 0)
+ }
+ 
+-func TestProcessMessageShouldProcessLogTypeFunction(t *testing.T) {
++func countMetricsByName(samples []metrics.MetricSample, name string) int {
++	count := 0
++	for _, sample := range samples {
++		if sample.Name == name {
++			count++
++		}
++	}
++	return count
++}
++
++func platformReportJSON(requestID, status string, memoryUsed, memorySize float64) []byte {
++	return []byte(fmt.Sprintf(`{
++		"time": %q,
++		"type": "platform.report",
++		"record": {
++			"requestId": %q,
++			"status": %q,
++			"metrics": {
++				"maxMemoryUsedMB": %.1f,
++				"memorySizeMB": %.1f
++			}
++		}
++	}`, time.Unix(0, 0).UTC().Format(logMessageTimeLayout), requestID, status, memoryUsed, memorySize))
++}
++
++func unmarshalPlatformReport(t *testing.T, raw []byte) *LambdaLogAPIMessage {
++	t.Helper()
++	message := &LambdaLogAPIMessage{}
++	require.NoError(t, json.Unmarshal(raw, message))
++	return message
++}
++
++func newOOMTestCollector(t *testing.T, demux aggregator.Demultiplexer, requestID string) *LambdaLogsCollector {
++	t.Helper()
++
++	executionContext := &executioncontext.ExecutionContext{}
++	return newOOMTestCollectorWithExecutionContext(t, demux, requestID, executionContext)
++}
++
++func newOOMTestCollectorWithExecutionContext(t *testing.T, demux aggregator.Demultiplexer, requestID string, executionContext *executioncontext.ExecutionContext) *LambdaLogsCollector {
++	t.Helper()
++
++	arn := "arn:aws:lambda:us-east-1:123456789012:function:test-function"
++	metricTags := []string{"functionname:test-function"}
++	tags := Tags{Tags: metricTags}
++	executionContext.SetFromInvocation(arn, requestID)
++
++	lc := NewLambdaLogCollector(
++		make(chan<- *config.ChannelMessage),
++		demux,
++		&tags,
++		true,
++		true,
++		executionContext,
++		func() {},
++		make(chan<- float64),
++	)
++	lc.lastRequestID = requestID
++	return lc
++}
++
++func processAndWaitForSamples(lc *LambdaLogsCollector, message LambdaLogAPIMessage, expectedSamples int) ([]metrics.MetricSample, []metrics.MetricSample) {
++	lc.Start()
++	lc.In <- []LambdaLogAPIMessage{message}
++	return lc.demux.(*aggregator.TestAgentDemultiplexer).WaitForNumberOfSamples(expectedSamples, 0, 100*time.Millisecond)
++}
++
++func TestProcessMessageShouldProcessLogTypeFunctionOutOfMemory(t *testing.T) {
+ 	demux := aggregator.InitTestAgentDemultiplexerWithFlushInterval(time.Hour)
+ 	defer demux.Stop(false)
+ 	message := &LambdaLogAPIMessage{
+@@ -413,14 +481,170 @@ func TestProcessMessageShouldProcessLogTypeFunction(t *testing.T) {
+ 	mockExecutionContext.SetFromInvocation(arn, lastRequestID)
+ 
+ 	lc := NewLambdaLogCollector(make(chan<- *config.ChannelMessage), demux, &tags, true, computeEnhancedMetrics, mockExecutionContext, func() {}, make(chan<- float64))
++	lc.lastRequestID = lastRequestID
+ 
+-	go lc.processMessage(message)
++	received, timed := processAndWaitForSamples(lc, *message, 2)
+ 
+-	received, timed := demux.WaitForNumberOfSamples(2, 0, 100*time.Millisecond)
+ 	assert.Len(t, received, 2)
+ 	assert.Len(t, timed, 0)
+-	assert.Equal(t, serverlessMetrics.OutOfMemoryMetric, received[0].Name)
+-	assert.Equal(t, serverlessMetrics.ErrorsMetric, received[1].Name)
++	assert.Equal(t, 1, countMetricsByName(received, serverlessMetrics.OutOfMemoryMetric))
++	assert.Equal(t, 1, countMetricsByName(received, serverlessMetrics.ErrorsMetric))
++}
++
++func TestProcessPlatformReportOOMEmitsEnhancedMetricsFromJSON(t *testing.T) {
++	demux := aggregator.InitTestAgentDemultiplexerWithFlushInterval(time.Hour)
++	defer demux.Stop(false)
++
++	requestID := "8286a188-ba32-4475-8077-530cd35c09a9"
++	message := unmarshalPlatformReport(t, platformReportJSON(requestID, "error", 512, 512))
++	lc := newOOMTestCollector(t, demux, requestID)
++
++	received, timed := processAndWaitForSamples(lc, *message, 8)
++	assert.Len(t, timed, 0)
++	assert.Equal(t, 1, countMetricsByName(received, serverlessMetrics.OutOfMemoryMetric))
++	assert.Equal(t, 1, countMetricsByName(received, serverlessMetrics.ErrorsMetric))
++}
++
++func TestProcessPlatformReportOOMAlsoEmitsWhenMemoryExceedsLimit(t *testing.T) {
++	demux := aggregator.InitTestAgentDemultiplexerWithFlushInterval(time.Hour)
++	defer demux.Stop(false)
++
++	requestID := "5d607cb8-d752-43e7-8f99-f5b5f6a17452"
++	message := unmarshalPlatformReport(t, platformReportJSON(requestID, "error", 768, 512))
++	lc := newOOMTestCollector(t, demux, requestID)
++
++	received, timed := processAndWaitForSamples(lc, *message, 8)
++	assert.Len(t, timed, 0)
++	assert.Equal(t, 1, countMetricsByName(received, serverlessMetrics.OutOfMemoryMetric))
++	assert.Equal(t, 1, countMetricsByName(received, serverlessMetrics.ErrorsMetric))
++}
++
++func TestProcessPlatformReportOOMRequiresQualifyingReport(t *testing.T) {
++	testCases := []struct {
++		name       string
++		status     string
++		memoryUsed float64
++		memorySize float64
++	}{
++		{
++			name:       "non-error status at memory limit",
++			status:     "success",
++			memoryUsed: 256,
++			memorySize: 256,
++		},
++		{
++			name:       "error status below memory limit",
++			status:     "error",
++			memoryUsed: 255,
++			memorySize: 256,
++		},
++		{
++			name:       "error status with zero memory used",
++			status:     "error",
++			memoryUsed: 0,
++			memorySize: 256,
++		},
++	}
++
++	for _, tc := range testCases {
++		t.Run(tc.name, func(t *testing.T) {
++			demux := aggregator.InitTestAgentDemultiplexerWithFlushInterval(time.Hour)
++			defer demux.Stop(false)
++
++			requestID := "negative-" + strings.ReplaceAll(tc.name, " ", "-")
++			message := unmarshalPlatformReport(t, platformReportJSON(requestID, tc.status, tc.memoryUsed, tc.memorySize))
++			lc := newOOMTestCollector(t, demux, requestID)
++
++			received, timed := processAndWaitForSamples(lc, *message, 6)
++			assert.Len(t, timed, 0)
++			assert.Equal(t, 0, countMetricsByName(received, serverlessMetrics.OutOfMemoryMetric))
++		})
++	}
++}
++
++func TestOutOfMemoryEnhancedMetricsAreDedupedPerRequest(t *testing.T) {
++	demux := aggregator.InitTestAgentDemultiplexerWithFlushInterval(time.Hour)
++	defer demux.Stop(false)
++
++	requestID := "fa7d55f7-4054-412f-9567-8f607c8376db"
++	lc := newOOMTestCollector(t, demux, requestID)
++	functionOOM := &LambdaLogAPIMessage{
++		logType:      logTypeFunction,
++		time:         time.Now(),
++		stringRecord: "JavaScript heap out of memory",
++	}
++	reportOOM := unmarshalPlatformReport(t, platformReportJSON(requestID, "error", 1024, 1024))
++
++	lc.Start()
++	lc.In <- []LambdaLogAPIMessage{*functionOOM, *reportOOM}
++
++	received, timed := demux.WaitForNumberOfSamples(8, 0, 100*time.Millisecond)
++	assert.Len(t, timed, 0)
++	assert.Equal(t, 1, countMetricsByName(received, serverlessMetrics.OutOfMemoryMetric))
++}
++
++func TestPersistedOOMDeduplicationPreventsDuplicateAfterRestore(t *testing.T) {
++	demux := aggregator.InitTestAgentDemultiplexerWithFlushInterval(time.Hour)
++	defer demux.Stop(false)
++
++	const stateFile = "/tmp/dd-lambda-extension-cache.json"
++	require.NoError(t, os.RemoveAll(stateFile))
++	t.Cleanup(func() {
++		_ = os.Remove(stateFile)
++	})
++
++	requestID := "72946812-35dd-4382-b180-92f51cb7168c"
++	lc := newOOMTestCollector(t, demux, requestID)
++	firstReport := unmarshalPlatformReport(t, platformReportJSON(requestID, "error", 384, 256))
++
++	received, timed := processAndWaitForSamples(lc, *firstReport, 8)
++	require.Len(t, timed, 0)
++	require.Equal(t, 1, countMetricsByName(received, serverlessMetrics.OutOfMemoryMetric))
++
++	require.NoError(t, lc.executionContext.SaveCurrentExecutionContext())
++
++	restoredCollector := newOOMTestCollector(t, demux, requestID)
++	require.NoError(t, restoredCollector.executionContext.RestoreCurrentStateFromFile())
++	replayedReport := unmarshalPlatformReport(t, platformReportJSON(requestID, "error", 384, 256))
++
++	restoredCollector.Start()
++	restoredCollector.In <- []LambdaLogAPIMessage{*replayedReport}
++
++	replayed, replayedTimed := demux.WaitForNumberOfSamples(14, 0, 100*time.Millisecond)
++	assert.Len(t, replayedTimed, 0)
++	assert.Equal(t, 1, countMetricsByName(replayed, serverlessMetrics.OutOfMemoryMetric))
++}
++
++func TestPersistedOOMDeduplicationPreventsDuplicateForArbitraryRequest(t *testing.T) {
++	demux := aggregator.InitTestAgentDemultiplexerWithFlushInterval(time.Hour)
++	defer demux.Stop(false)
++
++	const stateFile = "/tmp/dd-lambda-extension-cache.json"
++	require.NoError(t, os.RemoveAll(stateFile))
++	t.Cleanup(func() {
++		_ = os.Remove(stateFile)
++	})
++
++	requestID := "2b11979d-d2f0-4e1a-8840-5357f2c20b42"
++	lc := newOOMTestCollector(t, demux, requestID)
++	firstReport := unmarshalPlatformReport(t, platformReportJSON(requestID, "error", 640, 512))
++
++	received, timed := processAndWaitForSamples(lc, *firstReport, 8)
++	require.Len(t, timed, 0)
++	require.Equal(t, 1, countMetricsByName(received, serverlessMetrics.OutOfMemoryMetric))
++
++	require.NoError(t, lc.executionContext.SaveCurrentExecutionContext())
++
++	restoredCollector := newOOMTestCollector(t, demux, requestID)
++	require.NoError(t, restoredCollector.executionContext.RestoreCurrentStateFromFile())
++	replayedReport := unmarshalPlatformReport(t, platformReportJSON(requestID, "error", 640, 512))
++
++	restoredCollector.Start()
++	restoredCollector.In <- []LambdaLogAPIMessage{*replayedReport}
++
++	replayed, replayedTimed := demux.WaitForNumberOfSamples(14, 0, 100*time.Millisecond)
++	assert.Len(t, replayedTimed, 0)
++	assert.Equal(t, 1, countMetricsByName(replayed, serverlessMetrics.OutOfMemoryMetric))
+ }
+ 
+ func TestProcessLogMessageLogsEnabled(t *testing.T) {
+diff --git a/mimo_test_command.sh b/mimo_test_command.sh
+new file mode 100755
+--- /dev/null
++++ b/mimo_test_command.sh
+@@ -0,0 +1,20 @@
++#!/bin/bash
++set -uo pipefail
++cd /testbed
++
++
++cd /testbed
++
++
++case "${1:-}" in
++  base)
++    go test -v -count=1 ./pkg/serverless/executioncontext/... ./pkg/serverless/logs/...
++    ;;
++  new)
++    go test -v -count=1 ./pkg/serverless/logs/... -run '^(TestProcessMessageShouldProcessLogTypeFunctionOutOfMemory|TestProcessPlatformReportOOMEmitsEnhancedMetricsFromJSON|TestProcessPlatformReportOOMAlsoEmitsWhenMemoryExceedsLimit|TestProcessPlatformReportOOMRequiresQualifyingReport|TestOutOfMemoryEnhancedMetricsAreDedupedPerRequest|TestPersistedOOMDeduplicationPreventsDuplicateAfterRestore|TestPersistedOOMDeduplicationPreventsDuplicateForArbitraryRequest)$'
++    ;;
++  *)
++    echo "usage: $0 base|new"
++    exit 2
++    ;;
++esac

tasks/format-code-task-000056/tests/test.sh

@@ -0,0 +1,38 @@
+#!/bin/bash
+# Code grading, mimoagent opensource_code.py: reset the files the hidden test patch touches, apply it, run the
+# task's test command; reward 1 when it exits 0. A testbed failure (reset or apply) writes no reward: not scored.
+M=/var/lib/mimo
+CWD=/testbed
+V=/logs/verifier
+mkdir -p "$V"
+BASE=$(cat "$M/base" 2>/dev/null) || { echo "setup never ran (no $M/base)" >&2; exit 1; }
+cd "$CWD" || exit 1
+if [ -d "$M/git-hidden" ]; then rm -rf "$CWD/.git"; mv "$M/git-hidden" "$CWD/.git"; fi
+# the agent's change against the base, so its own commits count too
+git add -A >/dev/null 2>&1 && git -c core.fileMode=false diff --cached "$BASE" > "$V/agent.diff"
+git reset -q
+if ! (
+while IFS= read -r tf; do
+  [ -z "$tf" ] && continue
+  if git cat-file -e $BASE:"$tf" 2>/dev/null; then
+    git checkout $BASE -- "$tf" 2>/dev/null || true
+  else
+    git rm -f --cached "$tf" >/dev/null 2>&1 || true
+    rm -f "$tf"
+  fi
+done <<'EOF_RESET_TEST_FILES'
+pkg/serverless/logs/logs_test.go
+mimo_test_command.sh
+EOF_RESET_TEST_FILES
+
+); then echo "could not reset the test files (testbed problem, not scored)" >&2; exit 1; fi
+if ! git apply --verbose /tests/test.patch > "$V/apply.log" 2>&1; then
+  cat "$V/apply.log" >&2
+  echo "the hidden tests could not be applied (testbed problem, not scored)" >&2
+  exit 1
+fi
+timeout 1800 sh -c "$(cat /tests/test_command.sh)" > "$V/test_output.log" 2>&1
+RC=$?
+tail -c 12000 "$V/test_output.log"
+if [ $RC -eq 0 ]; then echo 1 > "$V/reward.txt"; else echo 0 > "$V/reward.txt"; fi
+echo "test command exited $RC"

tasks/format-code-task-000056/tests/test_command.sh

@@ -0,0 +1 @@
+bash /testbed/mimo_test_command.sh base && bash /testbed/mimo_test_command.sh new

tasks/format-code-task-000058/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:91b7266714fc6d0a66653e4b15a81820625e6019b8510e625560989946c22c47

tasks/format-code-task-000058/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com

tasks/format-code-task-000058/environment/setup/setup.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# Runs once, as root, before the agent starts: Harbor's [environment.healthcheck] unpacks this folder into
+# /var/lib/mimo and runs it. Generated by mimo_harbor; see its README for how each step maps to Xiaomi's harness.
+M=/var/lib/mimo
+[ -f "$M/ready" ] && exit 0
+chmod 700 "$M"
+fail() { echo "setup: $*" >&2; exit 1; }
+# the answer-leak blocklist (mimoagent), applied to /etc/hosts by the agent right after it installs
+write_blocklist() { cp "$M/files/blocklist" "$M/blocklist"; }
+
+# Code: mimoagent environments/datasets/opensource_code.py + base.py, as the explorer runs it (app/runner/domains.py Code).
+CWD=/testbed
+git config --global --add safe.directory "$CWD"
+cd "$CWD" || fail "no working directory $CWD"
+if ! git rev-parse --git-dir >/dev/null 2>&1; then   # a bare source tree: give it a baseline commit
+  git init -q && git add -A && git -c user.name=mimo -c user.email=mimo@localhost commit -q -m baseline --allow-empty
+fi
+BASE=$(git rev-parse HEAD 2>/dev/null)
+[ ${#BASE} -eq 40 ] || fail "could not resolve the base commit in $CWD"
+echo "$BASE" > "$M/base"
+# Images are built with history truncated at the base. If one is not, the fix could be read out of git log,
+# so .git is hidden while the agent works and put back for grading.
+LATER=$(git rev-list --all --not "$BASE" 2>/dev/null | head -n 5 | grep -c . || true)
+rm -f /tmp/fail.log /tmp/pass.log /tmp/patch.diff /tmp/test_patch.diff /tmp/*.log 2>/dev/null; rm -rf /tmp/claude-0 /tmp/claude-* /tmp/testem-* /tmp/puppeteer_dev_chrome_profile-* 2>/dev/null; rm -f /tmp/test_files.json 2>/dev/null; rm -rf /tmp/jest_* /tmp/jest-* /tmp/build /tmp/build_env 2>/dev/null; rm -rf /tmp/pytest-of-root /tmp/pytest-* /tmp/.pytest_cache /tmp/__pycache__ 2>/dev/null; rm -rf /tmp/go-build* /tmp/standards /tmp/wordpress /tmp/zig-* /tmp/testbase /tmp/testbed 2>/dev/null; rm -f /tmp/tmp*.tmp /tmp/*.bak /tmp/expect* /tmp/butwas* 2>/dev/null; rm -rf /root/.cache/go-build 2>/dev/null; rm -rf /tests 2>/dev/null; find /logs -mindepth 1 -delete 2>/dev/null; mkdir -p /logs/agent /logs/verifier /logs/artifacts; chmod 777 /logs/agent /logs/verifier /logs/artifacts; find /var/log -type f -delete 2>/dev/null || true; rm -rf /var/lib/postgresql/*/*/log /var/lib/postgresql/*/*/pg_log 2>/dev/null || true; find /var/lib/mysql /var/lib/mongodb -type f -name '*.log' -delete 2>/dev/null || true; rm -rf /root/.npm/_logs /root/.babel.json /root/.pytest_cache 2>/dev/null || true; find / -maxdepth 4 -xdev -name task_description.md -path '*/.build_env/*' -delete 2>/dev/null || true; true
+git clean -fdx --exclude=node_modules --exclude=bower_components --exclude=.husky --exclude=vendor --exclude=third_party --exclude=_deps --exclude=vcpkg_installed --exclude=.venv --exclude=venv --exclude=.gradle --exclude=target --exclude=.build --exclude=lib --exclude=.bundle --exclude=Manifest.toml --exclude=_build --exclude=.stack-work --exclude=dist-newstyle >/dev/null 2>&1
+find /root/.m2 -type f \( -name '*-SNAPSHOT.jar' -o -name '*-SNAPSHOT-sources.jar' -o -name '*-SNAPSHOT-tests.jar' -o -name '*-SNAPSHOT-test-sources.jar' \) -delete 2>/dev/null; rm -rf /root/.julia/compiled 2>/dev/null; rm -rf /root/.gradle/caches/build-cache-* /root/.gradle/daemon 2>/dev/null; rm -rf /root/.gradle/caches/*/scripts /root/.gradle/caches/jars-* 2>/dev/null; rm -rf /root/.cache/bazel 2>/dev/null; true
+if [ "$LATER" -gt 0 ]; then
+  mv "$CWD/.git" "$M/git-hidden"
+  echo "history not truncated at ${BASE:0:12}: .git hidden while the agent works"
+fi
+write_blocklist
+
+touch "$M/ready"
+echo "setup done"

tasks/format-code-task-000058/tests/test.patch

@@ -0,0 +1,114 @@
+diff --git a/rtloader/test/datadog_agent/datadog_agent.go b/rtloader/test/datadog_agent/datadog_agent.go
+index e0f202c0b34e5b..ef8474ab6dd84c 100644
+--- a/rtloader/test/datadog_agent/datadog_agent.go
++++ b/rtloader/test/datadog_agent/datadog_agent.go
+@@ -4,6 +4,7 @@ import (
+ 	"fmt"
+ 	"io/ioutil"
+ 	"os"
++	"time"
+ 	"runtime"
+ 	"strings"
+ 	"unsafe"
+@@ -33,6 +34,7 @@ extern void writePersistentCache(char*, char*);
+ extern char* readPersistentCache(char*);
+ extern char* obfuscateSQL(char*, char*, char**);
+ extern char* obfuscateSQLExecPlan(char*, bool, char**);
++extern double getProcessStartTime();
+ 
+ 
+ static void initDatadogAgentTests(rtloader_t *rtloader) {
+@@ -50,6 +52,7 @@ static void initDatadogAgentTests(rtloader_t *rtloader) {
+    set_read_persistent_cache_cb(rtloader, readPersistentCache);
+    set_obfuscate_sql_cb(rtloader, obfuscateSQL);
+    set_obfuscate_sql_exec_plan_cb(rtloader, obfuscateSQLExecPlan);
++   set_get_process_start_time_cb(rtloader, getProcessStartTime);
+ }
+ */
+ import "C"
+@@ -282,3 +285,10 @@ func obfuscateSQLExecPlan(rawQuery *C.char, normalize C.bool, errResult **C.char
+ 		return nil
+ 	}
+ }
++
++var processStartTime = float64(time.Now().Unix())
++
++//export getProcessStartTime
++func getProcessStartTime() float64 {
++	return processStartTime
++}
+diff --git a/rtloader/test/datadog_agent/datadog_agent_test.go b/rtloader/test/datadog_agent/datadog_agent_test.go
+index 6f17ebfe12174d..beb88133b4f1e5 100644
+--- a/rtloader/test/datadog_agent/datadog_agent_test.go
++++ b/rtloader/test/datadog_agent/datadog_agent_test.go
+@@ -4,6 +4,7 @@ import (
+ 	"fmt"
+ 	"os"
+ 	"regexp"
++	"strconv"
+ 	"testing"
+ 
+ 	"github.com/DataDog/datadog-agent/rtloader/test/helpers"
+@@ -578,3 +579,34 @@ func TestObfuscateSqlExecPlanErrors(t *testing.T) {
+ 
+ 	helpers.AssertMemoryUsage(t)
+ }
++
++func TestProcessStartTime(t *testing.T) {
++	// Reset memory counters
++	helpers.ResetMemoryStats()
++
++	exp := processStartTime // copy the value before anything touches it
++
++	code := fmt.Sprintf(`
++	with open(r'%s', 'w') as f:
++		pst = datadog_agent.get_process_start_time()
++		assert type(pst) == type(0.0)
++		f.write(pst.hex())
++	`, tmpfile.Name())
++
++	out, err := run(code)
++	if err != nil {
++		t.Fatal(err)
++	}
++
++	val, err := strconv.ParseFloat(out, 64)
++	if err != nil {
++		t.Fatal(err)
++	}
++
++	if val != exp {
++		t.Errorf("Unexpected printed value: '%s'", out)
++	}
++
++	// Check for leaks
++	helpers.AssertMemoryUsage(t)
++}
+diff --git a/mimo_build_env.tar.gz.b64 b/mimo_build_env.tar.gz.b64
+new file mode 100644
+--- /dev/null
++++ b/mimo_build_env.tar.gz.b64
+@@ -0,0 +1,12 @@
++H4sIAAAAAAAAA+2UbU/bMBDH+zqf4tRNGwg5IX0Y0qRO6sOASoUiYJqQJlWubVqvjh1sh6n79HPS
++0RUY8AY0bbufFNk6X+7+9j9OPC2k4hOhr5PaS7Eb2Gu3qzFwd6zmabuRNhqtNG21artp2t5r16D9
++Yoo2KJynFqBmjfGP5T21/pcS//LfC+cnzGQZ1Tx28+fr8YT/wezm2v/dVhr8b71LmzXYfT4JD/Of
+++884VMZPBY+iV3D2TXo2B2/goN+HtAlb1dgGRgsnHDCjL5VkXuoZ8cs8RIS1xjqQGhbFVBReqpht
++R1xcE7eqNWMsFLpd/GTp50ZDM04bm6l5FW6u4uGFoQ7mKBXSL7pHI9iy4qqQVnCYLsF6ZSgX9q0D
++523QA2XvUoj0kir5nXpp9HZ0U5NkkMs8LK9KEjK1gi6IWzovMpJTtqCzsB1yFVQsaaag8SEJ0hJd
++KFVq6ZX3ZN0V+kl/ZwfcnJZylJxaaqVwkc2A2Mt1WlLdrihbcGmB5HfjjN+LZHQhgAwGw7Nub/Rx
++cnJxfjg+bnTGxyHY+zQcDSaDj0fjznh/H+I4CSLfpNHqna+vt3RuDdteBYPk00KDnwvg1FNuZpOw
++Qe2hdNtFG74naw1lIOofjCf9/VH34KxTJ8P7SVIzVfDQ8Ddr5d8j+Eo+a0O4yK1g1AsepkxRWxni
++6vClajEa3PQY3a9THUbi51aIBB5OWLtB1M89kmqPZHNBlR3DuY2GvdPu6cXkpHt+2Kk/VTJ5/7iq
++sujMVIcZPi3DO+EB4unMQb3KqANhptC+kwK5hji5ZUISx3H0p68+giAIgiAIgiAIgiAIgiAIgiAI
++giAIgiD/CD8AmGc2XwAoAAA=
+diff --git a/mimo_test_command.sh b/mimo_test_command.sh
+new file mode 100755
+--- /dev/null
++++ b/mimo_test_command.sh
+@@ -0,0 +1,6 @@
++#!/usr/bin/env bash
++# Verifier for the opensource-code dataset. Exit 0 means resolved.
++set -uo pipefail
++cd /testbed || exit 1
++base64 -d mimo_build_env.tar.gz.b64 | tar -xzf - -C /testbed || exit 1
++exec bash -lc 'cd /testbed && bash -eo pipefail /testbed/.build_env/test_command.sh'

tasks/format-code-task-000058/tests/test.sh

@@ -0,0 +1,40 @@
+#!/bin/bash
+# Code grading, mimoagent opensource_code.py: reset the files the hidden test patch touches, apply it, run the
+# task's test command; reward 1 when it exits 0. A testbed failure (reset or apply) writes no reward: not scored.
+M=/var/lib/mimo
+CWD=/testbed
+V=/logs/verifier
+mkdir -p "$V"
+BASE=$(cat "$M/base" 2>/dev/null) || { echo "setup never ran (no $M/base)" >&2; exit 1; }
+cd "$CWD" || exit 1
+if [ -d "$M/git-hidden" ]; then rm -rf "$CWD/.git"; mv "$M/git-hidden" "$CWD/.git"; fi
+# the agent's change against the base, so its own commits count too
+git add -A >/dev/null 2>&1 && git -c core.fileMode=false diff --cached "$BASE" > "$V/agent.diff"
+git reset -q
+if ! (
+while IFS= read -r tf; do
+  [ -z "$tf" ] && continue
+  if git cat-file -e $BASE:"$tf" 2>/dev/null; then
+    git checkout $BASE -- "$tf" 2>/dev/null || true
+  else
+    git rm -f --cached "$tf" >/dev/null 2>&1 || true
+    rm -f "$tf"
+  fi
+done <<'EOF_RESET_TEST_FILES'
+rtloader/test/datadog_agent/datadog_agent.go
+rtloader/test/datadog_agent/datadog_agent_test.go
+mimo_build_env.tar.gz.b64
+mimo_test_command.sh
+EOF_RESET_TEST_FILES
+
+); then echo "could not reset the test files (testbed problem, not scored)" >&2; exit 1; fi
+if ! git apply --verbose /tests/test.patch > "$V/apply.log" 2>&1; then
+  cat "$V/apply.log" >&2
+  echo "the hidden tests could not be applied (testbed problem, not scored)" >&2
+  exit 1
+fi
+timeout 1800 sh -c "$(cat /tests/test_command.sh)" > "$V/test_output.log" 2>&1
+RC=$?
+tail -c 12000 "$V/test_output.log"
+if [ $RC -eq 0 ]; then echo 1 > "$V/reward.txt"; else echo 0 > "$V/reward.txt"; fi
+echo "test command exited $RC"

tasks/format-code-task-000058/tests/test_command.sh

@@ -0,0 +1 @@
+bash /testbed/mimo_test_command.sh

tasks/format-code-task-000060/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:c3b5d988c97689ca6a49626f265b52954d31b1110062a6df7f52ec16e8f8cb6f

tasks/format-code-task-000060/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com

tasks/format-code-task-000060/environment/setup/setup.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# Runs once, as root, before the agent starts: Harbor's [environment.healthcheck] unpacks this folder into
+# /var/lib/mimo and runs it. Generated by mimo_harbor; see its README for how each step maps to Xiaomi's harness.
+M=/var/lib/mimo
+[ -f "$M/ready" ] && exit 0
+chmod 700 "$M"
+fail() { echo "setup: $*" >&2; exit 1; }
+# the answer-leak blocklist (mimoagent), applied to /etc/hosts by the agent right after it installs
+write_blocklist() { cp "$M/files/blocklist" "$M/blocklist"; }
+
+# Code: mimoagent environments/datasets/opensource_code.py + base.py, as the explorer runs it (app/runner/domains.py Code).
+CWD=/testbed
+git config --global --add safe.directory "$CWD"
+cd "$CWD" || fail "no working directory $CWD"
+if ! git rev-parse --git-dir >/dev/null 2>&1; then   # a bare source tree: give it a baseline commit
+  git init -q && git add -A && git -c user.name=mimo -c user.email=mimo@localhost commit -q -m baseline --allow-empty
+fi
+BASE=$(git rev-parse HEAD 2>/dev/null)
+[ ${#BASE} -eq 40 ] || fail "could not resolve the base commit in $CWD"
+echo "$BASE" > "$M/base"
+# Images are built with history truncated at the base. If one is not, the fix could be read out of git log,
+# so .git is hidden while the agent works and put back for grading.
+LATER=$(git rev-list --all --not "$BASE" 2>/dev/null | head -n 5 | grep -c . || true)
+rm -f /tmp/fail.log /tmp/pass.log /tmp/patch.diff /tmp/test_patch.diff /tmp/*.log 2>/dev/null; rm -rf /tmp/claude-0 /tmp/claude-* /tmp/testem-* /tmp/puppeteer_dev_chrome_profile-* 2>/dev/null; rm -f /tmp/test_files.json 2>/dev/null; rm -rf /tmp/jest_* /tmp/jest-* /tmp/build /tmp/build_env 2>/dev/null; rm -rf /tmp/pytest-of-root /tmp/pytest-* /tmp/.pytest_cache /tmp/__pycache__ 2>/dev/null; rm -rf /tmp/go-build* /tmp/standards /tmp/wordpress /tmp/zig-* /tmp/testbase /tmp/testbed 2>/dev/null; rm -f /tmp/tmp*.tmp /tmp/*.bak /tmp/expect* /tmp/butwas* 2>/dev/null; rm -rf /root/.cache/go-build 2>/dev/null; rm -rf /tests 2>/dev/null; find /logs -mindepth 1 -delete 2>/dev/null; mkdir -p /logs/agent /logs/verifier /logs/artifacts; chmod 777 /logs/agent /logs/verifier /logs/artifacts; find /var/log -type f -delete 2>/dev/null || true; rm -rf /var/lib/postgresql/*/*/log /var/lib/postgresql/*/*/pg_log 2>/dev/null || true; find /var/lib/mysql /var/lib/mongodb -type f -name '*.log' -delete 2>/dev/null || true; rm -rf /root/.npm/_logs /root/.babel.json /root/.pytest_cache 2>/dev/null || true; find / -maxdepth 4 -xdev -name task_description.md -path '*/.build_env/*' -delete 2>/dev/null || true; true
+git clean -fdx --exclude=node_modules --exclude=bower_components --exclude=.husky --exclude=vendor --exclude=third_party --exclude=_deps --exclude=vcpkg_installed --exclude=.venv --exclude=venv --exclude=.gradle --exclude=target --exclude=.build --exclude=lib --exclude=.bundle --exclude=Manifest.toml --exclude=_build --exclude=.stack-work --exclude=dist-newstyle >/dev/null 2>&1
+find /root/.m2 -type f \( -name '*-SNAPSHOT.jar' -o -name '*-SNAPSHOT-sources.jar' -o -name '*-SNAPSHOT-tests.jar' -o -name '*-SNAPSHOT-test-sources.jar' \) -delete 2>/dev/null; rm -rf /root/.julia/compiled 2>/dev/null; rm -rf /root/.gradle/caches/build-cache-* /root/.gradle/daemon 2>/dev/null; rm -rf /root/.gradle/caches/*/scripts /root/.gradle/caches/jars-* 2>/dev/null; rm -rf /root/.cache/bazel 2>/dev/null; true
+if [ "$LATER" -gt 0 ]; then
+  mv "$CWD/.git" "$M/git-hidden"
+  echo "history not truncated at ${BASE:0:12}: .git hidden while the agent works"
+fi
+write_blocklist
+
+touch "$M/ready"
+echo "setup done"

tasks/format-code-task-000060/tests/test.patch

@@ -0,0 +1,140 @@
+diff --git a/internal/controller/datadogmonitor/monitor_test.go b/internal/controller/datadogmonitor/monitor_test.go
+index ae74b3500..da8a4bb66 100644
+--- a/internal/controller/datadogmonitor/monitor_test.go
++++ b/internal/controller/datadogmonitor/monitor_test.go
+@@ -18,32 +18,17 @@ import (
+ 	"time"
+ 
+ 	"github.com/stretchr/testify/assert"
++	"k8s.io/utils/ptr"
+ 
+ 	datadogapi "github.com/DataDog/datadog-api-client-go/v2/api/datadog"
+-	datadogV1 "github.com/DataDog/datadog-api-client-go/v2/api/datadogV1"
++	"github.com/DataDog/datadog-api-client-go/v2/api/datadogV1"
++
+ 	datadoghqv1alpha1 "github.com/DataDog/datadog-operator/api/datadoghq/v1alpha1"
+ )
+ 
+ const dateFormat = "2006-01-02 15:04:05.999999999 -0700 MST"
+ 
+ func Test_buildMonitor(t *testing.T) {
+-	evalDelay := int64(100)
+-	escalationMsg := "This is an escalation message"
+-	valTrue := true
+-	newGroupDelay := int64(400)
+-	noDataTimeframe := int64(15)
+-	renotifyInterval := int64(1440)
+-	renotifyOccurrences := int64(1)
+-	renotifyStatuses := []datadogV1.MonitorRenotifyStatusType{
+-		datadogV1.MONITORRENOTIFYSTATUSTYPE_ALERT,
+-		datadogV1.MONITORRENOTIFYSTATUSTYPE_WARN,
+-	}
+-
+-	timeoutH := int64(2)
+-	critThreshold := "0.05"
+-	warnThreshold := "0.02"
+-	priority := int64(3)
+-
+ 	dm := &datadoghqv1alpha1.DatadogMonitor{
+ 		Spec: datadoghqv1alpha1.DatadogMonitorSpec{
+ 			RestrictedRoles: []string{"an-admin-uuid"},
+@@ -51,36 +36,46 @@ func Test_buildMonitor(t *testing.T) {
+ 			Type:            "metric alert",
+ 			Name:            "Test monitor",
+ 			Message:         "Something went wrong",
+-			Priority:        priority,
++			Priority:        int64(3),
+ 			Tags: []string{
+ 				"env:staging",
+ 				"kube_namespace:test",
+ 				"kube_cluster:test.staging",
+ 			},
+ 			Options: datadoghqv1alpha1.DatadogMonitorOptions{
+-				EnableLogsSample:       &valTrue,
+-				EvaluationDelay:        &evalDelay,
+-				EscalationMessage:      &escalationMsg,
+-				IncludeTags:            &valTrue,
+-				GroupbySimpleMonitor:   &valTrue,
+-				Locked:                 &valTrue,
+-				NewGroupDelay:          &newGroupDelay,
+-				NoDataTimeframe:        &noDataTimeframe,
++				EnableLogsSample:       ptr.To(true),
++				EvaluationDelay:        ptr.To(int64(100)),
++				EscalationMessage:      ptr.To("This is an escalation message"),
++				IncludeTags:            ptr.To(true),
++				GroupbySimpleMonitor:   ptr.To(true),
++				Locked:                 ptr.To(true),
++				NewGroupDelay:          ptr.To(int64(400)),
++				NoDataTimeframe:        ptr.To(int64(15)),
+ 				NotificationPresetName: "show_all",
+ 				NotifyBy: []string{
+ 					"env",
+ 					"kube_namespace",
+ 					"kube_cluster",
+ 				},
+-				NotifyNoData:        &valTrue,
++				NotifyNoData:        ptr.To(true),
+ 				OnMissingData:       "default",
+-				RenotifyOccurrences: &renotifyOccurrences,
+-				RenotifyInterval:    &renotifyInterval,
+-				RenotifyStatuses:    renotifyStatuses,
+-				TimeoutH:            &timeoutH,
++				RenotifyOccurrences: ptr.To(int64(1)),
++				RenotifyInterval:    ptr.To(int64(1440)),
++				RenotifyStatuses: []datadogV1.MonitorRenotifyStatusType{
++					datadogV1.MONITORRENOTIFYSTATUSTYPE_ALERT,
++					datadogV1.MONITORRENOTIFYSTATUSTYPE_WARN,
++				},
++				SchedulingOptions: &datadoghqv1alpha1.DatadogMonitorOptionsSchedulingOptions{
++					EvaluationWindow: &datadoghqv1alpha1.DatadogMonitorOptionsSchedulingOptionsEvaluationWindow{
++						DayStarts:   ptr.To("01:00"),
++						HourStarts:  ptr.To(int32(2)),
++						MonthStarts: ptr.To(int32(3)),
++					},
++				},
++				TimeoutH: ptr.To(int64(2)),
+ 				Thresholds: &datadoghqv1alpha1.DatadogMonitorOptionsThresholds{
+-					Critical: &critThreshold,
+-					Warning:  &warnThreshold,
++					Critical: ptr.To("0.05"),
++					Warning:  ptr.To("0.02"),
+ 				},
+ 			},
+ 		},
+@@ -151,6 +146,15 @@ func Test_buildMonitor(t *testing.T) {
+ 	assert.Equal(t, dm.Spec.Options.RenotifyStatuses, monitor.Options.GetRenotifyStatuses(), "discrepancy found in parameter: RenotifyStatuses")
+ 	assert.Equal(t, dm.Spec.Options.RenotifyStatuses, monitorUR.Options.GetRenotifyStatuses(), "discrepancy found in parameter: RenotifyStatuses")
+ 
++	assert.Equal(t, *dm.Spec.Options.SchedulingOptions.EvaluationWindow.DayStarts, monitor.Options.SchedulingOptions.EvaluationWindow.GetDayStarts(), "discrepancy found in parameter: EvaluationWindow.DayStarts")
++	assert.Equal(t, *dm.Spec.Options.SchedulingOptions.EvaluationWindow.DayStarts, monitorUR.Options.SchedulingOptions.EvaluationWindow.GetDayStarts(), "discrepancy found in parameter: EvaluationWindow.DayStarts")
++
++	assert.Equal(t, *dm.Spec.Options.SchedulingOptions.EvaluationWindow.HourStarts, monitor.Options.SchedulingOptions.EvaluationWindow.GetHourStarts(), "discrepancy found in parameter: EvaluationWindow.HourStarts")
++	assert.Equal(t, *dm.Spec.Options.SchedulingOptions.EvaluationWindow.HourStarts, monitorUR.Options.SchedulingOptions.EvaluationWindow.GetHourStarts(), "discrepancy found in parameter: EvaluationWindow.HourStarts")
++
++	assert.Equal(t, *dm.Spec.Options.SchedulingOptions.EvaluationWindow.MonthStarts, monitor.Options.SchedulingOptions.EvaluationWindow.GetMonthStarts(), "discrepancy found in parameter: EvaluationWindow.MonthStarts")
++	assert.Equal(t, *dm.Spec.Options.SchedulingOptions.EvaluationWindow.MonthStarts, monitorUR.Options.SchedulingOptions.EvaluationWindow.GetMonthStarts(), "discrepancy found in parameter: EvaluationWindow.MonthStarts")
++
+ 	assert.Equal(t, *dm.Spec.Options.TimeoutH, monitor.Options.GetTimeoutH(), "discrepancy found in parameter: TimeoutH")
+ 	assert.Equal(t, *dm.Spec.Options.TimeoutH, monitorUR.Options.GetTimeoutH(), "discrepancy found in parameter: TimeoutH")
+ 
+diff --git a/mimo_build_env.tar.gz.b64 b/mimo_build_env.tar.gz.b64
+new file mode 100644
+--- /dev/null
++++ b/mimo_build_env.tar.gz.b64
+@@ -0,0 +1,5 @@
++H4sIAAAAAAAAA+3STWrDMBAFYK1zirmArZ9G8SpHyK77oFgiMdgakMc5f2MX2lJos0koTd63mUES
++aMRTfZi6Pu5TPmt1L+ai8X6pF9/r0lvvrDPO+vVGGWt90yjyd5voi2mUUIhUYZbfzl3b/6fqz/wl
++jbJveRhCjvV4ut0d1/K3L/4jf+fm/Ncb4xWZ243wsyfPv420BH9IcXVkmluqzlS1PGXZWqp1lyWV
++HHrdcpbCfZ+KjkFC5OPAuRMumqoyZXqd/8/ynXbv66u/fhwAAAAAAAAAAAAAAAAAAAAAwIN7A7fv
++5lEAKAAA
+diff --git a/mimo_test_command.sh b/mimo_test_command.sh
+new file mode 100755
+--- /dev/null
++++ b/mimo_test_command.sh
+@@ -0,0 +1,6 @@
++#!/usr/bin/env bash
++
++set -uo pipefail
++cd /testbed || exit 1
++base64 -d mimo_build_env.tar.gz.b64 | tar -xzf - -C /testbed || exit 1
++exec bash -lc 'cd /testbed && bash -eo pipefail /testbed/.build_env/test_command.sh'

tasks/format-code-task-000060/tests/test.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+# Code grading, mimoagent opensource_code.py: reset the files the hidden test patch touches, apply it, run the
+# task's test command; reward 1 when it exits 0. A testbed failure (reset or apply) writes no reward: not scored.
+M=/var/lib/mimo
+CWD=/testbed
+V=/logs/verifier
+mkdir -p "$V"
+BASE=$(cat "$M/base" 2>/dev/null) || { echo "setup never ran (no $M/base)" >&2; exit 1; }
+cd "$CWD" || exit 1
+if [ -d "$M/git-hidden" ]; then rm -rf "$CWD/.git"; mv "$M/git-hidden" "$CWD/.git"; fi
+# the agent's change against the base, so its own commits count too
+git add -A >/dev/null 2>&1 && git -c core.fileMode=false diff --cached "$BASE" > "$V/agent.diff"
+git reset -q
+if ! (
+while IFS= read -r tf; do
+  [ -z "$tf" ] && continue
+  if git cat-file -e $BASE:"$tf" 2>/dev/null; then
+    git checkout $BASE -- "$tf" 2>/dev/null || true
+  else
+    git rm -f --cached "$tf" >/dev/null 2>&1 || true
+    rm -f "$tf"
+  fi
+done <<'EOF_RESET_TEST_FILES'
+internal/controller/datadogmonitor/monitor_test.go
+mimo_build_env.tar.gz.b64
+mimo_test_command.sh
+EOF_RESET_TEST_FILES
+
+); then echo "could not reset the test files (testbed problem, not scored)" >&2; exit 1; fi
+if ! git apply --verbose /tests/test.patch > "$V/apply.log" 2>&1; then
+  cat "$V/apply.log" >&2
+  echo "the hidden tests could not be applied (testbed problem, not scored)" >&2
+  exit 1
+fi
+timeout 1800 sh -c "$(cat /tests/test_command.sh)" > "$V/test_output.log" 2>&1
+RC=$?
+tail -c 12000 "$V/test_output.log"
+if [ $RC -eq 0 ]; then echo 1 > "$V/reward.txt"; else echo 0 > "$V/reward.txt"; fi
+echo "test command exited $RC"

tasks/format-code-task-000060/tests/test_command.sh

@@ -0,0 +1 @@
+bash /testbed/mimo_test_command.sh

tasks/format-code-task-000067/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:5987f5e3ecbdc9568af6bd08225a63fb8c4daba8bb466709a64be3a195ed15e6

tasks/format-code-task-000067/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com

tasks/format-code-task-000067/environment/setup/setup.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# Runs once, as root, before the agent starts: Harbor's [environment.healthcheck] unpacks this folder into
+# /var/lib/mimo and runs it. Generated by mimo_harbor; see its README for how each step maps to Xiaomi's harness.
+M=/var/lib/mimo
+[ -f "$M/ready" ] && exit 0
+chmod 700 "$M"
+fail() { echo "setup: $*" >&2; exit 1; }
+# the answer-leak blocklist (mimoagent), applied to /etc/hosts by the agent right after it installs
+write_blocklist() { cp "$M/files/blocklist" "$M/blocklist"; }
+
+# Code: mimoagent environments/datasets/opensource_code.py + base.py, as the explorer runs it (app/runner/domains.py Code).
+CWD=/testbed
+git config --global --add safe.directory "$CWD"
+cd "$CWD" || fail "no working directory $CWD"
+if ! git rev-parse --git-dir >/dev/null 2>&1; then   # a bare source tree: give it a baseline commit
+  git init -q && git add -A && git -c user.name=mimo -c user.email=mimo@localhost commit -q -m baseline --allow-empty
+fi
+BASE=$(git rev-parse HEAD 2>/dev/null)
+[ ${#BASE} -eq 40 ] || fail "could not resolve the base commit in $CWD"
+echo "$BASE" > "$M/base"
+# Images are built with history truncated at the base. If one is not, the fix could be read out of git log,
+# so .git is hidden while the agent works and put back for grading.
+LATER=$(git rev-list --all --not "$BASE" 2>/dev/null | head -n 5 | grep -c . || true)
+rm -f /tmp/fail.log /tmp/pass.log /tmp/patch.diff /tmp/test_patch.diff /tmp/*.log 2>/dev/null; rm -rf /tmp/claude-0 /tmp/claude-* /tmp/testem-* /tmp/puppeteer_dev_chrome_profile-* 2>/dev/null; rm -f /tmp/test_files.json 2>/dev/null; rm -rf /tmp/jest_* /tmp/jest-* /tmp/build /tmp/build_env 2>/dev/null; rm -rf /tmp/pytest-of-root /tmp/pytest-* /tmp/.pytest_cache /tmp/__pycache__ 2>/dev/null; rm -rf /tmp/go-build* /tmp/standards /tmp/wordpress /tmp/zig-* /tmp/testbase /tmp/testbed 2>/dev/null; rm -f /tmp/tmp*.tmp /tmp/*.bak /tmp/expect* /tmp/butwas* 2>/dev/null; rm -rf /root/.cache/go-build 2>/dev/null; rm -rf /tests 2>/dev/null; find /logs -mindepth 1 -delete 2>/dev/null; mkdir -p /logs/agent /logs/verifier /logs/artifacts; chmod 777 /logs/agent /logs/verifier /logs/artifacts; find /var/log -type f -delete 2>/dev/null || true; rm -rf /var/lib/postgresql/*/*/log /var/lib/postgresql/*/*/pg_log 2>/dev/null || true; find /var/lib/mysql /var/lib/mongodb -type f -name '*.log' -delete 2>/dev/null || true; rm -rf /root/.npm/_logs /root/.babel.json /root/.pytest_cache 2>/dev/null || true; find / -maxdepth 4 -xdev -name task_description.md -path '*/.build_env/*' -delete 2>/dev/null || true; true
+git clean -fdx --exclude=node_modules --exclude=bower_components --exclude=.husky --exclude=vendor --exclude=third_party --exclude=_deps --exclude=vcpkg_installed --exclude=.venv --exclude=venv --exclude=.gradle --exclude=target --exclude=.build --exclude=lib --exclude=.bundle --exclude=Manifest.toml --exclude=_build --exclude=.stack-work --exclude=dist-newstyle >/dev/null 2>&1
+find /root/.m2 -type f \( -name '*-SNAPSHOT.jar' -o -name '*-SNAPSHOT-sources.jar' -o -name '*-SNAPSHOT-tests.jar' -o -name '*-SNAPSHOT-test-sources.jar' \) -delete 2>/dev/null; rm -rf /root/.julia/compiled 2>/dev/null; rm -rf /root/.gradle/caches/build-cache-* /root/.gradle/daemon 2>/dev/null; rm -rf /root/.gradle/caches/*/scripts /root/.gradle/caches/jars-* 2>/dev/null; rm -rf /root/.cache/bazel 2>/dev/null; true
+if [ "$LATER" -gt 0 ]; then
+  mv "$CWD/.git" "$M/git-hidden"
+  echo "history not truncated at ${BASE:0:12}: .git hidden while the agent works"
+fi
+write_blocklist
+
+touch "$M/ready"
+echo "setup done"

tasks/format-code-task-000067/tests/test.patch

@@ -0,0 +1,87 @@
+diff --git a/directory/tests/test_directory.py b/directory/tests/test_directory.py
+index 3b655e109394d..927eaccb4ce7c 100644
+--- a/directory/tests/test_directory.py
++++ b/directory/tests/test_directory.py
+@@ -331,29 +331,44 @@ def test_non_existent_directory_ignore_missing(aggregator):
+     aggregator.assert_service_check('system.disk.directory.exists', DirectoryCheck.WARNING, tags=expected_tags)
+ 
+ 
+-def test_os_error_mid_walk_emits_error_and_continues(aggregator, monkeypatch, caplog):
++def test_os_error_mid_walk_emits_error_and_continues(aggregator, caplog):
+     caplog.set_level(logging.WARNING)
+ 
+-    def mock_walk(folder, *args, **kwargs):
+-        from datadog_checks.directory.traverse import walk
+-
+-        walker = walk(folder, *args, **kwargs)
+-        yield next(walker)
+-        raise OSError('Permission denied')
+-
+-    monkeypatch.setattr('datadog_checks.directory.directory.walk', mock_walk)
++
++
++
++
++
++
++
++
++
++
+ 
+     with temp_directory() as tdir:
+-
+-        # Create folder
+-        mkdir(os.path.join(tdir, 'a_folder'))
++
++        os.makedirs(os.path.join(tdir, 'bad_folder_a'), mode=0o377)
++        os.makedirs(os.path.join(tdir, 'bad_folder_b'), mode=0o377)
++
++        os.makedirs(os.path.join(tdir, 'ok'))
++        with open(os.path.join(tdir, 'ok', 'file'), 'w') as f:
++            f.write('')
+ 
+         # Run Check
+         instance = {'directory': tdir, 'recursive': True}
+         check = DirectoryCheck('directory', {}, [instance])
+         check.check(instance)
+ 
+-    assert 'Permission denied' in caplog.text
++
++        os.chmod(os.path.join(tdir, 'bad_folder_a'), 0o777)
++        os.chmod(os.path.join(tdir, 'bad_folder_b'), 0o777)
++
++    aggregator.assert_metric("system.disk.directory.files", count=1, value=1)
++
++    permission_denied_log_lines = [line for line in caplog.text.splitlines() if 'Permission denied' in line]
++    assert len(permission_denied_log_lines) == 2
++    assert 'bad_folder_a' in caplog.text
++    assert 'bad_folder_b' in caplog.text
+ 
+ 
+ def test_no_recursive_symlink_loop(aggregator):
+diff --git a/mimo_build_env.tar.gz.b64 b/mimo_build_env.tar.gz.b64
+new file mode 100644
+--- /dev/null
++++ b/mimo_build_env.tar.gz.b64
+@@ -0,0 +1,9 @@
++H4sIAAAAAAAAA+2UzW7TQBDHffZTDKkUgSrbccDpjQvijvoC1np3QlbYu2Y/Qi3KuzPrKEkbCXpp
++hVDnd/Bas/O5/7XLLupetWj2VfZSrIibpplX4nKd3+tmXa9X9aZZr7NVXTc3TQbNi3X0gOiDcACZ
++szb8ze+p/f+U8qx/QB9aaYdBGFX63fPVeEL/un5fn/T/sCF7Tc9NBqvna+HPvHL9r95U0buq06ai
++OwCd8LvcY4ACo4VRj7gVus+lgvl6dKjy/Ao+ORQBIVmiRwd6C8YGED3Z1QSjQ48m5FpBEc9ey4+V
++wn1lYt/D/T0km1DkMpxcUu7PxkeHsE/daA9CSvRedz2Vs1TFFEmIOdjncjdYSnAL4trdnTqsyhSc
++b7VRFzYowjQiUAjeoYRD+Bz78xdcp+q30UDYHUaj4v7c/dtj7XfQoRRkOzm21rfonHXtoFX7Q/Tf
++Whx0OBrpc6LPygRtInqqkULA4feo6ZxAPB4qTfmVzn/xBd2gaXJrQKHRqBYwp/NAFgoMTsuAChRl
++kcE6jb7MfYTi3HLhYRY2iQqFhIW30Um8OJTZRcig90nT5RIeiF0ds09pY5zCjoqTYPSWhphP6fDf
++ODmW4wTFfpH/64vNMAzDMAzDMAzDMAzDMAzDMAzDMAzDMAzzSvgNIXRQMAAoAAA=
+diff --git a/mimo_test_command.sh b/mimo_test_command.sh
+new file mode 100755
+--- /dev/null
++++ b/mimo_test_command.sh
+@@ -0,0 +1,6 @@
++#!/usr/bin/env bash
++
++set -uo pipefail
++cd /testbed || exit 1
++base64 -d mimo_build_env.tar.gz.b64 | tar -xzf - -C /testbed || exit 1
++exec bash -lc 'cd /testbed && bash -eo pipefail /testbed/.build_env/test_command.sh'

tasks/format-code-task-000067/tests/test.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+# Code grading, mimoagent opensource_code.py: reset the files the hidden test patch touches, apply it, run the
+# task's test command; reward 1 when it exits 0. A testbed failure (reset or apply) writes no reward: not scored.
+M=/var/lib/mimo
+CWD=/testbed
+V=/logs/verifier
+mkdir -p "$V"
+BASE=$(cat "$M/base" 2>/dev/null) || { echo "setup never ran (no $M/base)" >&2; exit 1; }
+cd "$CWD" || exit 1
+if [ -d "$M/git-hidden" ]; then rm -rf "$CWD/.git"; mv "$M/git-hidden" "$CWD/.git"; fi
+# the agent's change against the base, so its own commits count too
+git add -A >/dev/null 2>&1 && git -c core.fileMode=false diff --cached "$BASE" > "$V/agent.diff"
+git reset -q
+if ! (
+while IFS= read -r tf; do
+  [ -z "$tf" ] && continue
+  if git cat-file -e $BASE:"$tf" 2>/dev/null; then
+    git checkout $BASE -- "$tf" 2>/dev/null || true
+  else
+    git rm -f --cached "$tf" >/dev/null 2>&1 || true
+    rm -f "$tf"
+  fi
+done <<'EOF_RESET_TEST_FILES'
+directory/tests/test_directory.py
+mimo_build_env.tar.gz.b64
+mimo_test_command.sh
+EOF_RESET_TEST_FILES
+
+); then echo "could not reset the test files (testbed problem, not scored)" >&2; exit 1; fi
+if ! git apply --verbose /tests/test.patch > "$V/apply.log" 2>&1; then
+  cat "$V/apply.log" >&2
+  echo "the hidden tests could not be applied (testbed problem, not scored)" >&2
+  exit 1
+fi
+timeout 1800 sh -c "$(cat /tests/test_command.sh)" > "$V/test_output.log" 2>&1
+RC=$?
+tail -c 12000 "$V/test_output.log"
+if [ $RC -eq 0 ]; then echo 1 > "$V/reward.txt"; else echo 0 > "$V/reward.txt"; fi
+echo "test command exited $RC"

tasks/format-code-task-000067/tests/test_command.sh

@@ -0,0 +1 @@
+bash /testbed/mimo_test_command.sh

tasks/format-code-task-000069/environment/Dockerfile

@@ -0,0 +1,3 @@
+# The task's image, pinned by digest (the same as [environment].docker_image).
+# Setup runs at start from [environment.healthcheck]; environment/setup/ is its readable copy.
+FROM docker.io/xiaomimimo/mimo-v2.6-rl-oss@sha256:25f55976fd5a55e4280cb4ec3c6e7273e7f58d066037d5282d9428d38cafa679

tasks/format-code-task-000069/environment/setup/files/blocklist

@@ -0,0 +1,33 @@
+# mimo answer-leak blocklist (mimoagent)
+0.0.0.0 github.com
+0.0.0.0 www.github.com
+0.0.0.0 api.github.com
+0.0.0.0 raw.githubusercontent.com
+0.0.0.0 codeload.github.com
+0.0.0.0 objects.githubusercontent.com
+0.0.0.0 gist.github.com
+0.0.0.0 gist.githubusercontent.com
+0.0.0.0 gitlab.com
+0.0.0.0 bitbucket.org
+0.0.0.0 sourceforge.net
+0.0.0.0 git.kernel.org
+0.0.0.0 chromium.googlesource.com
+0.0.0.0 android.googlesource.com
+0.0.0.0 sourcegraph.com
+0.0.0.0 bugs.chromium.org
+0.0.0.0 issues.oss-fuzz.com
+0.0.0.0 oss-fuzz.com
+0.0.0.0 osv.dev
+0.0.0.0 api.osv.dev
+0.0.0.0 bugzilla.mozilla.org
+0.0.0.0 proxy.golang.org
+0.0.0.0 sum.golang.org
+0.0.0.0 pkg.go.dev
+0.0.0.0 google.com
+0.0.0.0 www.google.com
+0.0.0.0 bing.com
+0.0.0.0 www.bing.com
+0.0.0.0 duckduckgo.com
+0.0.0.0 html.duckduckgo.com
+0.0.0.0 search.brave.com
+0.0.0.0 stackoverflow.com