maestro-flow 0.4.6 → 0.4.7

package/.codex/skills/team-coordinate/specs/role-catalog.md

@@ -0,0 +1,132 @@
+# Role Catalog — Team Coordinate
+
+Dynamic role reference for CSV wave execution. Coordinator uses this catalog during Phase 1 (task analysis) to detect capabilities, assign roles, compute waves, and generate evaluation criteria.
+
+## 1. Signal Detection Table
+
+Scan task description keywords to infer capabilities. Multi-match allowed.
+
+| Signal | Keywords | Role | Prefix | Responsibility |
+|--------|----------|------|--------|---------------|
+| Research | investigate, explore, compare, survey, find, research, discover, benchmark, study | researcher | RESEARCH | exploration |
+| Writing | write, draft, document, article, report, blog, describe, explain, summarize, content | writer | DRAFT | code-gen (docs) |
+| Coding | implement, build, code, fix, refactor, develop, create app, program, migrate, port | developer | IMPL | code-gen (code) |
+| Design | design, architect, plan, structure, blueprint, model, schema, wireframe, layout | designer | DESIGN | orchestration |
+| Analysis | analyze, review, audit, assess, evaluate, inspect, examine, diagnose, profile | analyst | ANALYSIS | read-only |
+| Testing | test, verify, validate, QA, quality, check, assert, coverage, regression | tester | TEST | validation |
+| Planning | plan, breakdown, organize, schedule, decompose, roadmap, strategy, prioritize | planner | PLAN | orchestration |
+
+**No match**: Default to single `general` role with `TASK` prefix.
+
+**Max 5 roles** per session. Merge overlapping capabilities if exceeded.
+
+## 2. Wave Tier Mapping
+
+| Tier | Wave | Roles | Description |
+|------|------|-------|-------------|
+| 0 | 1 | researcher, planner | Knowledge gathering (always executes) |
+| 1 | 2 | designer, analyst | Design + evaluation (evaluated) |
+| 2 | 3 | developer, writer | Creation (evaluated) |
+| 3 | 4 | tester, analyst | Validation (evaluated) |
+
+**general** role: assigned to wave matching its task's natural tier.
+
+## 3. Evaluation Criteria Templates
+
+Wave 1 tasks always get `"always"`. Wave 2+ tasks get conditional criteria:
+
+| Wave | Default Criteria | When to Skip |
+|------|-----------------|--------------|
+| 2 (Design) | `"if wave_1 findings indicate design or architecture needed"` | Research found no design needs |
+| 3 (Implement) | `"if task requires code changes or document creation"` | Pure analysis task, no implementation |
+| 4 (Validate) | `"if wave_3 produced testable artifacts (files_modified non-empty)"` | No artifacts to validate |
+
+Coordinator MAY customize criteria per-task based on task description specifics.
+
+## 4. Role Behavior Guide
+
+Each role adopts a specific perspective when executing tasks. This section is referenced by the `instruction` parameter of `spawn_agents_on_csv`.
+
+| Role | Perspective | Typical Output |
+|------|------------|----------------|
+| researcher | Systematic investigation, evidence-driven, hypothesis testing | Research findings, exploration notes |
+| developer | Implementation-focused, code modification, testing | Source files, implementation summaries |
+| analyst | Multi-dimensional evaluation, scoring, gap identification | Analysis reports, severity matrices |
+| designer | Architecture, structure, interface, data model design | Design specs, architecture docs |
+| tester | Validation, test creation, regression, coverage analysis | Test results, coverage reports |
+| planner | Decomposition, sequencing, risk assessment, prioritization | Execution plans, task breakdowns |
+| writer | Documentation, content creation, clarity, consistency | Written documents, guides |
+| general | Adapt to task requirements | Task-specific deliverables |
+
+## 5. Behavioral Traits (Quality Contract)
+
+All agents — regardless of dynamically assigned role — MUST follow these traits. Embedded into the shared `instruction` parameter.
+
+### Accuracy
+
+- Files claimed as **created** → Read to confirm file exists and has content
+- Files claimed as **modified** → Read to confirm content actually changed
+- Analysis claimed as **complete** → verify deliverable exists
+
+### Quality Gate
+
+- Verify actual output (not planned output) before reporting
+- Verification fails → retry execution (max 2 retries)
+- Still fails → report `blocked` with details, NOT `completed`
+
+### Error Protocol
+
+- Primary approach fails → try alternative approach
+- 2 retries exhausted → report `failed` with evidence
+- NEVER skip verification and report completed
+
+## 6. Quality Scoring
+
+| Result | Score | Action |
+|--------|-------|--------|
+| Pass | >= 80% | result_status = completed |
+| Review | 60-79% | result_status = completed (with findings noting warnings) |
+| Fail | < 60% | Retry (max 2), then result_status = blocked |
+
+**Scoring Dimensions** (25% each):
+- **Completeness**: All required outputs present with substantive content
+- **Consistency**: Terminology, formatting, cross-references uniform
+- **Accuracy**: Factually correct and verifiable
+- **Depth**: Sufficient detail for downstream consumers
+
+## 7. Output Type Derivation
+
+| Task Signal | Output Type | Description |
+|-------------|-------------|-------------|
+| "write report", "analyze", "research" | artifact | New files in session directory |
+| "implement", "modify code", "fix bug" | codebase | Modify existing project files |
+| "implement feature + write summary" | mixed | Code changes + session artifact |
+
+## 8. Key File Inference
+
+For task decomposition, infer relevant files from capability and keywords:
+
+| Role | Strategy |
+|------|----------|
+| researcher | Domain keywords → likely directories (e.g., "auth" → `src/auth/**`) |
+| developer | Feature/module keywords → source files |
+| designer | Architecture keywords → config/schema files |
+| analyst | Target keywords → files under analysis |
+| tester | Test target → source + test files |
+| writer | Documentation target → relevant source for context |
+| planner | No specific files (abstract planning) |
+
+## 9. Complexity Scoring
+
+| Factor | Weight |
+|--------|--------|
+| Capability count | +1 each |
+| Cross-domain (3+ tiers) | +2 |
+| Parallel tracks | +1 each |
+| Serial depth | +1 per level |
+
+| Total | Level | Role Limit |
+|-------|-------|------------|
+| 1-3 | Low | 1-2 roles |
+| 4-6 | Medium | 2-3 roles |
+| 7+ | High | 3-5 roles |