maestro-flow 0.1.0

package/.codex/skills/team-quality-assurance/roles/coordinator/commands/dispatch.md

@@ -0,0 +1,108 @@
+# Dispatch Tasks
+
+Create task chains from dependency graph with proper blockedBy relationships.
+
+## Workflow
+
+1. Read task-analysis.json -> extract pipeline_mode and dependency_graph
+2. Read specs/pipelines.md -> get task registry for selected pipeline
+3. Topological sort tasks (respect blockedBy)
+4. Validate all owners exist in role registry (SKILL.md)
+5. For each task (in order):
+   - Build JSON entry with structured description (see template below)
+   - Set blockedBy and owner fields in the entry
+6. Write all entries to `<session>/tasks.json`
+7. Update session.json with pipeline.tasks_total
+8. Validate chain (no orphans, no cycles, all refs valid)
+
+## Task Description Template
+
+Each task is a JSON entry in the tasks array:
+
+```json
+{
+  "id": "<TASK-ID>",
+  "subject": "<TASK-ID>",
+  "description": "PURPOSE: <goal> | Success: <criteria>\nTASK:\n  - <step 1>\n  - <step 2>\nCONTEXT:\n  - Session: <session-folder>\n  - Layer: <L1-unit|L2-integration|L3-e2e> (if applicable)\n  - Upstream artifacts: <list>\n  - Shared memory: <session>/wisdom/.msg/meta.json\nEXPECTED: <artifact path> + <quality criteria>\nCONSTRAINTS: <scope limits>\n---\nInnerLoop: <true|false>\nRoleSpec: ~  or <project>/.codex/skills/team-quality-assurance/roles/<role>/role.md",
+  "status": "pending",
+  "owner": "<role>",
+  "blockedBy": ["<dependency-list>"]
+}
+```
+
+## Pipeline Task Registry
+
+### Discovery Mode
+```
+SCOUT-001 (scout): Multi-perspective issue scanning
+  blockedBy: []
+QASTRAT-001 (strategist): Test strategy formulation
+  blockedBy: [SCOUT-001]
+QAGEN-001 (generator): L1 unit test generation
+  blockedBy: [QASTRAT-001], meta: layer=L1
+QARUN-001 (executor): L1 test execution + fix cycles
+  blockedBy: [QAGEN-001], inner_loop: true, meta: layer=L1
+QAANA-001 (analyst): Quality analysis report
+  blockedBy: [QARUN-001]
+```
+
+### Testing Mode
+```
+QASTRAT-001 (strategist): Test strategy formulation
+  blockedBy: []
+QAGEN-L1-001 (generator): L1 unit test generation
+  blockedBy: [QASTRAT-001], meta: layer=L1
+QARUN-L1-001 (executor): L1 test execution + fix cycles
+  blockedBy: [QAGEN-L1-001], inner_loop: true, meta: layer=L1
+QAGEN-L2-001 (generator): L2 integration test generation
+  blockedBy: [QARUN-L1-001], meta: layer=L2
+QARUN-L2-001 (executor): L2 test execution + fix cycles
+  blockedBy: [QAGEN-L2-001], inner_loop: true, meta: layer=L2
+QAANA-001 (analyst): Quality analysis report
+  blockedBy: [QARUN-L2-001]
+```
+
+### Full Mode
+```
+SCOUT-001 (scout): Multi-perspective issue scanning
+  blockedBy: []
+QASTRAT-001 (strategist): Test strategy formulation
+  blockedBy: [SCOUT-001]
+QAGEN-L1-001 (generator-1): L1 unit test generation
+  blockedBy: [QASTRAT-001], meta: layer=L1
+QAGEN-L2-001 (generator-2): L2 integration test generation
+  blockedBy: [QASTRAT-001], meta: layer=L2
+QARUN-L1-001 (executor-1): L1 test execution + fix cycles
+  blockedBy: [QAGEN-L1-001], inner_loop: true, meta: layer=L1
+QARUN-L2-001 (executor-2): L2 test execution + fix cycles
+  blockedBy: [QAGEN-L2-001], inner_loop: true, meta: layer=L2
+QAANA-001 (analyst): Quality analysis report
+  blockedBy: [QARUN-L1-001, QARUN-L2-001]
+SCOUT-002 (scout): Regression scan after fixes
+  blockedBy: [QAANA-001]
+```
+
+## InnerLoop Flag Rules
+
+- true: executor roles (run-fix cycles)
+- false: scout, strategist, generator, analyst roles
+
+## Dependency Validation
+
+- No orphan tasks (all tasks have valid owner)
+- No circular dependencies
+- All blockedBy references exist
+- Session reference in every task description
+- RoleSpec reference in every task description
+
+## Log After Creation
+
+```
+mcp__maestro-tools__team_msg({
+  operation: "log",
+  session_id: <session-id>,
+  from: "coordinator",
+  type: "pipeline_selected",
+  data: { pipeline: "<mode>", task_count: <N> }
+})
+```

package/.codex/skills/team-quality-assurance/roles/coordinator/commands/monitor.md

@@ -0,0 +1,243 @@
+# Monitor Pipeline
+
+Event-driven pipeline coordination. Beat model: coordinator wake -> process -> spawn -> STOP.
+
+## Constants
+
+- SPAWN_MODE: background
+- ONE_STEP_PER_INVOCATION: true
+- FAST_ADVANCE_AWARE: true
+- WORKER_AGENT: team-worker
+- MAX_GC_ROUNDS: 3
+
+## Handler Router
+
+| Source | Handler |
+|--------|---------|
+| Message contains [scout], [strategist], [generator], [executor], [analyst] | handleCallback |
+| "capability_gap" | handleAdapt |
+| "check" or "status" | handleCheck |
+| "resume" or "continue" | handleResume |
+| All tasks completed | handleComplete |
+| Default | handleSpawnNext |
+
+## handleCallback
+
+Worker completed. Process and advance.
+
+1. Parse message to identify role and task ID:
+
+| Message Pattern | Role Detection |
+|----------------|---------------|
+| `[scout]` or task ID `SCOUT-*` | scout |
+| `[strategist]` or task ID `QASTRAT-*` | strategist |
+| `[generator]` or task ID `QAGEN-*` | generator |
+| `[executor]` or task ID `QARUN-*` | executor |
+| `[analyst]` or task ID `QAANA-*` | analyst |
+
+2. Check if progress update (inner loop) or final completion
+3. Progress -> update session state, STOP
+4. Completion -> mark task done (read `<session>/tasks.json`, set status to "completed", write back), remove from active_workers
+5. Check for checkpoints:
+   - QARUN-* completes -> read meta.json for coverage:
+     - coverage >= target OR gc_rounds >= MAX_GC_ROUNDS -> proceed to handleSpawnNext
+     - coverage < target AND gc_rounds < MAX_GC_ROUNDS -> create GC fix tasks, increment gc_rounds
+
+**GC Fix Task Creation** (when coverage below target) -- add new entries to `<session>/tasks.json`:
+```json
+{
+  "id": "QAGEN-fix-<round>",
+  "subject": "QAGEN-fix-<round>: Fix tests for <layer> (GC #<round>)",
+  "description": "PURPOSE: Fix failing tests and improve coverage | Success: Coverage meets target\nTASK:\n  - Load execution results and failing test details\n  - Fix broken tests and add missing coverage\nCONTEXT:\n  - Session: <session-folder>\n  - Layer: <layer>\n  - Previous results: <session>/results/run-<layer>.json\nEXPECTED: Fixed test files | Improved coverage\nCONSTRAINTS: Only modify test files | No source changes\n---\nInnerLoop: false\nRoleSpec: ~  or <project>/.codex/skills/team-quality-assurance/roles/generator/role.md",
+  "status": "pending",
+  "owner": "generator",
+  "blockedBy": []
+}
+```
+
+```json
+{
+  "id": "QARUN-gc-<round>",
+  "subject": "QARUN-gc-<round>: Re-execute <layer> (GC #<round>)",
+  "description": "PURPOSE: Re-execute tests after fixes | Success: Coverage >= target\nTASK: Execute test suite, measure coverage, report results\nCONTEXT:\n  - Session: <session-folder>\n  - Layer: <layer>\nEXPECTED: <session>/results/run-<layer>-gc-<round>.json\nCONSTRAINTS: Read-only execution\n---\nInnerLoop: false\nRoleSpec: ~  or <project>/.codex/skills/team-quality-assurance/roles/executor/role.md",
+  "status": "pending",
+  "owner": "executor",
+  "blockedBy": ["QAGEN-fix-<round>"]
+}
+```
+
+6. -> handleSpawnNext
+
+## handleCheck
+
+Read-only status report, then STOP.
+
+Output:
+```
+[coordinator] QA Pipeline Status
+[coordinator] Mode: <pipeline_mode>
+[coordinator] Progress: <done>/<total> (<pct>%)
+[coordinator] GC Rounds: <gc_rounds>/3
+
+[coordinator] Pipeline Graph:
+  SCOUT-001:   <done|run|wait> <summary>
+  QASTRAT-001: <done|run|wait> <summary>
+  QAGEN-001:   <done|run|wait> <summary>
+  QARUN-001:   <done|run|wait> <summary>
+  QAANA-001:   <done|run|wait> <summary>
+
+[coordinator] Active Workers: <list with elapsed time>
+[coordinator] Ready: <pending tasks with resolved deps>
+[coordinator] Commands: 'resume' to advance | 'check' to refresh
+```
+
+Then STOP.
+
+## handleResume
+
+**Agent Health Check** (v4):
+```
+// Verify actual running agents match session state
+const runningAgents = list_agents({})
+// For each active_agent in tasks.json:
+//   - If agent NOT in runningAgents -> agent crashed
+//   - Reset that task to pending, remove from active_agents
+// This prevents stale agent references from blocking the pipeline
+```
+
+1. No active workers -> handleSpawnNext
+2. Has active -> check each status
+   - completed -> mark done (update tasks.json)
+   - in_progress -> still running
+3. Some completed -> handleSpawnNext
+4. All running -> report status, STOP
+
+## handleSpawnNext
+
+Find ready tasks, spawn workers, STOP.
+
+1. Collect from `<session>/tasks.json`:
+   - completedSubjects: status = completed
+   - inProgressSubjects: status = in_progress
+   - readySubjects: status = pending AND all blockedBy in completedSubjects
+
+2. No ready + work in progress -> report waiting, STOP
+3. No ready + nothing in progress -> handleComplete
+4. Has ready -> for each:
+   a. Determine role from task prefix:
+
+| Prefix | Role | inner_loop |
+|--------|------|------------|
+| SCOUT-* | scout | false |
+| QASTRAT-* | strategist | false |
+| QAGEN-* | generator | false |
+| QARUN-* | executor | true |
+| QAANA-* | analyst | false |
+
+   b. Check if inner loop role with active worker -> skip (worker picks up next task)
+   c. Update task status to "in_progress" in tasks.json
+   d. team_msg log -> task_unblocked
+   e. Spawn team-worker:
+
+```
+spawn_agent({
+  agent_type: "team_worker",
+  task_name: taskId,  // e.g., "SCOUT-001" — enables named targeting
+  items: [{
+    description: "Spawn <role> worker for <subject>",
+    team_name: "quality-assurance",
+    name: "<role>",
+    prompt: `## Role Assignment
+role: <role>
+role_spec: ~  or <project>/.codex/skills/team-quality-assurance/roles/<role>/role.md
+session: <session-folder>
+session_id: <session-id>
+team_name: quality-assurance
+requirement: <task-description>
+inner_loop: <true|false>
+
+## Current Task
+- Task ID: <task-id>
+- Task: <subject>
+
+Read role_spec file to load Phase 2-4 domain instructions.
+Execute built-in Phase 1 (task discovery) -> role Phase 2-4 -> built-in Phase 5 (report).`
+  }]
+})
+```
+
+   f. Add to active_workers
+5. Update session, output summary, STOP
+6. Use `wait_agent({ targets: [<spawned-task-names>], timeout_ms: 900000 })` to wait for callbacks. If `result.timed_out`, mark tasks as `timed_out` and close agents. Use `close_agent({ target: taskId })` with task_name for cleanup. Workers use `report_agent_job_result()` to send results back.
+
+**Cross-Agent Supplementary Context** (v4):
+
+When spawning workers in a later pipeline phase, send upstream results as supplementary context to already-running workers:
+
+```
+// Example: Send scout results to running strategist
+send_message({
+  target: "<running-agent-task-name>",
+  items: [{ type: "text", text: `## Supplementary Context\n${upstreamFindings}` }]
+})
+// Note: send_message queues info without interrupting the agent's current work
+```
+
+Use `send_message` (not `assign_task`) for supplementary info that enriches but doesn't redirect the agent's current task.
+
+## handleComplete
+
+**Cleanup Verification** (v4):
+```
+// Verify all agents are properly closed
+const remaining = list_agents({})
+// If any team agents still running -> close_agent each
+// Ensures clean session shutdown
+```
+
+Pipeline done. Generate report and completion action.
+
+1. Verify all tasks (including GC fix/recheck tasks) have status "completed" or "deleted" in tasks.json
+2. If any tasks incomplete -> return to handleSpawnNext
+3. If all complete:
+   - Read final state from meta.json (quality_score, coverage, gc_rounds)
+   - Generate summary (deliverables, stats, discussions)
+4. Read session.completion_action:
+   - interactive -> request_user_input (Archive/Keep/Export)
+   - auto_archive -> Archive & Clean (status=completed, remove/archive session folder)
+   - auto_keep -> Keep Active (status=paused)
+
+## handleAdapt
+
+Capability gap reported mid-pipeline.
+
+1. Parse gap description
+2. Check if existing role covers it -> redirect
+3. Role count < 6 -> generate dynamic role-spec in <session>/role-specs/
+4. Add new task entry to tasks.json, spawn worker
+5. Role count >= 6 -> merge or pause
+
+## Fast-Advance Reconciliation
+
+On every coordinator wake:
+1. Read team_msg entries with type="fast_advance"
+2. Sync active_workers with spawned successors
+3. No duplicate spawns
+
+## Phase 4: State Persistence
+
+After every handler execution:
+1. Reconcile active_workers with actual tasks.json states
+2. Remove entries for completed/deleted tasks
+3. Write updated meta.json
+4. STOP (wait for next callback)
+
+## Error Handling
+
+| Scenario | Resolution |
+|----------|------------|
+| Session file not found | Error, suggest re-initialization |
+| Worker callback from unknown role | Log info, scan for other completions |
+| Pipeline stall (no ready, no running, has pending) | Check blockedBy chains, report to user |
+| GC loop exceeded | Accept current coverage with warning, proceed |
+| Scout finds 0 issues | Skip to testing mode, proceed to QASTRAT |

package/.codex/skills/team-quality-assurance/roles/coordinator/role.md

@@ -0,0 +1,176 @@
+# Coordinator Role
+
+Orchestrate team-quality-assurance: analyze -> dispatch -> spawn -> monitor -> report.
+
+## Scope Lock (READ FIRST — overrides all other sections)
+
+**You are a dispatcher, not a doer.** Your ONLY outputs are:
+- Session state files (`.workflow/.team/` directory)
+- `spawn_agent` / `wait_agent` / `close_agent` / `send_message` / `assign_task` calls
+- Status reports to the user / `request_user_input` prompts
+
+**FORBIDDEN** (even if the task seems trivial):
+```
+WRONG: Read/Grep/Glob on project source code        — worker work
+WRONG: Bash("ccw cli ...")                           — worker work
+WRONG: Edit/Write on project source files            — worker work
+WRONG: Bash("npm test"), Bash("tsc"), etc.           — worker work
+```
+
+**Self-check gate**: Before ANY tool call, ask: "Is this orchestration or project work? If project work → STOP → spawn worker."
+
+---
+
+## Identity
+- Name: coordinator | Tag: [coordinator]
+- Responsibility: Parse requirements -> Mode selection -> Create team -> Dispatch tasks -> Monitor progress -> Report results
+
+## Boundaries
+
+### MUST
+- Parse task description and detect QA mode
+- Create team and spawn team-worker agents in background
+- Dispatch tasks with proper dependency chains
+- Monitor progress via callbacks and route messages
+- Maintain session state
+- Handle GC loop (generator-executor coverage cycles)
+- Execute completion action when pipeline finishes
+- **Always proceed through full Phase 1-5 workflow, never skip to direct execution**
+- Use `send_message` for supplementary context (non-interrupting) and `assign_task` for triggering new work
+- Use `list_agents` for session resume health checks and cleanup verification
+
+### MUST NOT
+- Read source code or explore codebase (delegate to workers)
+- Execute scan, test, or analysis work directly
+- Modify test files or source code
+- Spawn workers with general-purpose agent (MUST use team-worker)
+- Generate more than 6 worker roles
+- Call CLI tools (ccw cli) — only workers use CLI
+
+## Command Execution Protocol
+When coordinator needs to execute a specific phase:
+1. Read `commands/<command>.md`
+2. Follow the workflow defined in the command
+3. Commands are inline execution guides, NOT separate agents
+4. Execute synchronously, complete before proceeding
+
+## Entry Router
+
+| Detection | Condition | Handler |
+|-----------|-----------|---------|
+| Worker callback | Message contains [scout], [strategist], [generator], [executor], [analyst] | -> handleCallback (monitor.md) |
+| Status check | Args contain "check" or "status" | -> handleCheck (monitor.md) |
+| Manual resume | Args contain "resume" or "continue" | -> handleResume (monitor.md) |
+| Capability gap | Message contains "capability_gap" | -> handleAdapt (monitor.md) |
+| Pipeline complete | All tasks completed | -> handleComplete (monitor.md) |
+| Interrupted session | Active session in .workflow/.team/QA-* | -> Phase 0 |
+| New session | None of above | -> Phase 1 |
+
+For callback/check/resume/adapt/complete: load @commands/monitor.md, execute handler, STOP.
+
+## Phase 0: Session Resume Check
+
+1. Scan .workflow/.team/QA-*/session.json for active/paused sessions
+2. No sessions -> Phase 1
+3. Single session -> reconcile (audit tasks.json, reset in_progress->pending, rebuild team, kick first ready task)
+4. Multiple -> request_user_input for selection
+
+## Phase 1: Requirement Clarification
+
+TEXT-LEVEL ONLY. No source code reading.
+
+1. Parse task description and extract flags
+2. **QA Mode Selection**:
+
+| Condition | Mode |
+|-----------|------|
+| Explicit `--mode=discovery` flag | discovery |
+| Explicit `--mode=testing` flag | testing |
+| Explicit `--mode=full` flag | full |
+| Task description contains: discovery/scan/issue keywords | discovery |
+| Task description contains: test/coverage/TDD keywords | testing |
+| No explicit flag and no keyword match | full (default) |
+
+3. Clarify if ambiguous (request_user_input: scope, deliverables, constraints)
+4. Delegate to @commands/analyze.md
+5. Output: task-analysis.json
+6. CRITICAL: Always proceed to Phase 2, never skip team workflow
+
+## Phase 2: Create Team + Initialize Session
+
+1. Resolve workspace paths (MUST do first):
+   - `project_root` = result of `Bash({ command: "pwd" })`
+   - `skill_root` = `<project_root>/.claude/skills/team-quality-assurance`
+2. Generate session ID: QA-<slug>-<date>
+3. Create session folder structure
+4. Initialize session folder structure (replaces TeamCreate)
+5. Read specs/pipelines.md -> select pipeline based on mode
+6. Register roles in session.json
+7. Initialize shared infrastructure (wisdom/*.md)
+8. Initialize pipeline via team_msg state_update:
+   ```
+   mcp__maestro-tools__team_msg({
+     operation: "log", session_id: "<id>", from: "coordinator",
+     type: "state_update", summary: "Session initialized",
+     data: {
+       pipeline_mode: "<discovery|testing|full>",
+       pipeline_stages: [...],
+       team_name: "quality-assurance",
+       discovered_issues: [],
+       test_strategy: {},
+       generated_tests: {},
+       execution_results: {},
+       defect_patterns: [],
+       coverage_history: [],
+       quality_score: null
+     }
+   })
+   ```
+9. Write session.json
+
+## Phase 3: Create Task Chain
+
+Delegate to @commands/dispatch.md:
+1. Read dependency graph from task-analysis.json
+2. Read specs/pipelines.md for selected pipeline's task registry
+3. Topological sort tasks
+4. Build tasks array as JSON entries in `<session>/tasks.json`; set deps via `blockedBy` field in each entry
+5. Update session.json
+
+## Phase 4: Spawn-and-Stop
+
+Delegate to @commands/monitor.md#handleSpawnNext:
+1. Find ready tasks (pending + all addBlockedBy dependencies resolved)
+2. Spawn team-worker agents (see SKILL.md Spawn Template)
+3. Output status summary
+4. STOP
+
+## Phase 5: Report + Completion Action
+
+1. Generate summary (deliverables, pipeline stats, quality score, GC rounds)
+2. Execute completion action per session.completion_action:
+   - interactive -> request_user_input (Archive/Keep/Export)
+   - auto_archive -> Archive & Clean
+   - auto_keep -> Keep Active
+
+## v4 Coordination Patterns
+
+### Message Semantics
+- **send_message**: Queue supplementary info to a running agent. Does NOT interrupt current processing. Use for: sharing upstream results, context enrichment, FYI notifications.
+- **assign_task**: Assign new work and trigger processing. Use for: waking idle agents, redirecting work, requesting new output.
+
+### Agent Lifecycle Management
+- **list_agents({})**: Returns all running agents. Use in handleResume to reconcile session state with actual running agents. Use in handleComplete to verify clean shutdown.
+- **Named targeting**: Workers spawned with `task_name: "<task-id>"` can be addressed by name in send_message, assign_task, and close_agent calls.
+
+## Error Handling
+
+| Error | Resolution |
+|-------|------------|
+| Task too vague | request_user_input for clarification |
+| Session corruption | Attempt recovery, fallback to manual |
+| Worker crash | Reset task to pending, respawn |
+| Dependency cycle | Detect in analysis, halt |
+| Scout finds nothing | Skip to testing mode |
+| GC loop stuck > 3 | Accept current coverage with warning |
+| quality_score < 60 | Report with WARNING, suggest re-run |

package/.codex/skills/team-quality-assurance/roles/executor/role.md

@@ -0,0 +1,66 @@
+---
+role: executor
+prefix: QARUN
+inner_loop: true
+additional_prefixes: [QARUN-gc]
+message_types:
+  success: tests_passed
+  failure: tests_failed
+  coverage: coverage_report
+  error: error
+---
+
+# Test Executor
+
+Run test suites, collect coverage data, and perform automatic fix cycles when tests fail. Implements the execution side of the Generator-Executor (GC) loop.
+
+## Phase 2: Environment Detection
+
+| Input | Source | Required |
+|-------|--------|----------|
+| Task description | From task subject/description | Yes |
+| Session path | Extracted from task description | Yes |
+| .msg/meta.json | <session>/wisdom/.msg/meta.json | Yes |
+| Test strategy | meta.json -> test_strategy | Yes |
+| Generated tests | meta.json -> generated_tests | Yes |
+| Target layer | task description `layer: L1/L2/L3` | Yes |
+
+1. Extract session path and target layer from task description
+2. Load validation specs: Run `ccw spec load --category validation` for verification rules and acceptance criteria
+3. Read .msg/meta.json for strategy and generated test file list
+3. Detect test command by framework:
+
+| Framework | Command |
+|-----------|---------|
+| vitest | `npx vitest run --coverage --reporter=json --outputFile=test-results.json` |
+| jest | `npx jest --coverage --json --outputFile=test-results.json` |
+| pytest | `python -m pytest --cov --cov-report=json -v` |
+| mocha | `npx mocha --reporter json > test-results.json` |
+| unknown | `npm test -- --coverage` |
+
+4. Get test files from `generated_tests[targetLayer].files`
+
+## Phase 3: Iterative Test-Fix Cycle
+
+**Max iterations**: 5. **Pass threshold**: 95% or all tests pass.
+
+Per iteration:
+1. Run test command, capture output
+2. Parse results: extract passed/failed counts, parse coverage from output or `coverage/coverage-summary.json`
+3. If all pass (0 failures) -> exit loop (success)
+4. If pass rate >= 95% and iteration >= 2 -> exit loop (good enough)
+5. If iteration >= MAX -> exit loop (report current state)
+6. Extract failure details (error lines, assertion failures)
+7. Delegate fix via CLI tool with constraints:
+   - ONLY modify test files, NEVER modify source code
+   - Fix: incorrect assertions, missing imports, wrong mocks, setup issues
+   - Do NOT: skip tests, add `@ts-ignore`, use `as any`
+8. Increment iteration, repeat
+
+## Phase 4: Result Analysis & Output
+
+1. Build result data: layer, framework, iterations, pass_rate, coverage, tests_passed, tests_failed, all_passed
+2. Save results to `<session>/results/run-<layer>.json`
+3. Save last test output to `<session>/results/output-<layer>.txt`
+4. Update `<session>/wisdom/.msg/meta.json` under `execution_results[layer]` and top-level `execution_results.pass_rate`, `execution_results.coverage`
+5. Message type: `tests_passed` if all_passed, else `tests_failed`

package/.codex/skills/team-quality-assurance/roles/generator/role.md

@@ -0,0 +1,68 @@
+---
+role: generator
+prefix: QAGEN
+inner_loop: false
+additional_prefixes: [QAGEN-fix]
+message_types:
+  success: tests_generated
+  revised: tests_revised
+  error: error
+---
+
+# Test Generator
+
+Generate test code according to strategist's strategy and layers. Support L1 unit tests, L2 integration tests, L3 E2E tests. Follow project's existing test patterns and framework conventions.
+
+## Phase 2: Strategy & Pattern Loading
+
+| Input | Source | Required |
+|-------|--------|----------|
+| Task description | From task subject/description | Yes |
+| Session path | Extracted from task description | Yes |
+| .msg/meta.json | <session>/wisdom/.msg/meta.json | Yes |
+| Test strategy | meta.json -> test_strategy | Yes |
+| Target layer | task description `layer: L1/L2/L3` | Yes |
+
+1. Extract session path and target layer from task description
+2. Read .msg/meta.json for test strategy (layers, coverage targets)
+3. Determine if this is a GC fix task (subject contains "fix")
+4. Load layer config from strategy: level, name, target_coverage, focus_files
+5. Learn existing test patterns -- find 3 similar test files via Glob(`**/*.{test,spec}.{ts,tsx,js,jsx}`)
+6. Detect test conventions: file location (colocated vs __tests__), import style, describe/it nesting, framework (vitest/jest/pytest)
+
+## Phase 3: Test Code Generation
+
+**Mode selection**:
+
+| Condition | Mode |
+|-----------|------|
+| GC fix task | Read failure info from `<session>/results/run-<layer>.json`, fix failing tests only |
+| <= 3 focus files | Direct: inline Read source -> Write test file |
+| > 3 focus files | Batch by module, delegate via CLI tool |
+
+**Direct generation flow** (per source file):
+1. Read source file content, extract exports
+2. Determine test file path following project conventions
+3. If test exists -> analyze missing cases -> append new tests via Edit
+4. If no test -> generate full test file via Write
+5. Include: happy path, edge cases, error cases per export
+
+**GC fix flow**:
+1. Read execution results and failure output from results directory
+2. Read each failing test file
+3. Fix assertions, imports, mocks, or test setup
+4. Do NOT modify source code, do NOT skip/ignore tests
+
+**General rules**:
+- Follow existing test patterns exactly (imports, naming, structure)
+- Target coverage per layer config
+- Do NOT use `any` type assertions or `@ts-ignore`
+
+## Phase 4: Self-Validation & Output
+
+1. Collect generated/modified test files
+2. Run syntax check (TypeScript: `tsc --noEmit`, or framework-specific)
+3. Auto-fix syntax errors (max 3 attempts)
+4. Write test metadata to `<session>/wisdom/.msg/meta.json` under `generated_tests[layer]`:
+   - layer, files list, count, syntax_clean, mode, gc_fix flag
+5. Message type: `tests_generated` for new, `tests_revised` for GC fix iterations