opencode-workflow/skills/challenge-plan/SKILL.md

92 lines
2.7 KiB
Markdown

---
name: challenge-plan
description: "Silent audit and batch review of the implementation plan. Validates traceability, task granularity, dependency correctness, milestone quality, and scope discipline. Updates the single plan file in place."
---
Perform a silent, structured audit of the plan against the architecture and code design documents. Apply fixes directly to the plan file. Do not ask interactive questions.
**Announce at start:** "I'm using the challenge-plan skill to audit and review the implementation plan."
## Primary Input
- `docs/plan/{date}-{feature}.md`
- `docs/architecture/{date}-{feature}.md`
- `docs/code-design/{date}-{feature}.md`
## Primary Output (STRICT PATH)
- Updated `docs/plan/{date}-{feature}.md`
## Audit Phases
### Phase 1: Traceability
- Every task maps to upstream architecture or code design artifacts
- No orphan tasks exist
### Phase 2: Coverage
- All major implementation surfaces from upstream docs are represented
- Migration, integration, rollout, config, and observability tasks are included where needed
### Phase 3: Granularity
- Tasks are neither too vague nor too fine-grained
- Completion criteria are actionable
### Phase 4: Dependency Correctness
- Dependencies reflect actual implementation order
- Parallelization opportunities are valid
- No circular task graph exists
### Phase 5: Scope Discipline
- The plan does not introduce new design or product scope
- QA and Engineering duties are not mixed into planning incorrectly
### Phase 6: Structural Completeness
- All 12 required sections from the Output Format are present
- `# Open Questions` section exists (may be marked N/A with reason if no questions remain)
- Each task has defined dependencies, deliverables, and completion criteria
- The dependency graph matches the execution order text
## Review Output Format
Write the results directly into the `## Planning Review` section using this exact format:
```markdown
## Planning Review
### 1. Traceability & Coverage
- [PASS/FAIL] All required work is represented and traceable.
- Findings: ...
### 2. Task Quality
- [PASS/FAIL] Tasks are actionable and appropriately sized.
- Findings: ...
### 3. Dependency Logic
- [PASS/FAIL] Dependency order is correct and practical.
- Findings: ...
### 4. Scope Discipline
- [PASS/FAIL] No design drift or scope drift detected.
- Findings: ...
### 5. Structural Completeness
- [PASS/FAIL] All required sections present and Open Questions addressed.
- Findings: ...
### 6. Final Verdict
- [APPROVED/REJECTED]
- Reason: ...
```
## Guardrails
Do:
- Audit silently and fix issues in batch
- Keep the plan aligned to upstream design
- Use the exact review format
Do not:
- Ask interactive questions one-by-one
- Change architecture or code design
- Create new files