メニュー
Validate that implementation satisfies spec acceptance criteria. Maps each criterion in requirements.md to corresponding tests, runs them, and reports pass/fail per criterion. Use after implementation, before PR creation.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Create Conventional Commits with type prefix, breaking change detection, dry-run preview, and strict rules against ticket IDs in messages. **AUTOMATICALLY USE** when: (1) Creating git commits ("commit", "git commit") (2) Suggesting commit messages ("commit message") (3) Splitting commits ("split commits") (4) Fixing commit messages ("fix commit") (5) Any git commit related request
Orchestrate the full release process: push to develop, generate release summary, propose version bump, update package.json files, generate CHANGELOG, create release PR, and monitor CI. Supports dry-run preview mode. **AUTOMATICALLY USE** when: (1) Releasing a new version ("release", "cut a release", "prepare release") (2) Bumping version ("version bump", "bump version") (3) Creating release PRs ("release PR") (4) Generating changelogs ("changelog", "release notes") (5) Any release-related request
Record knowledge to the local knowledge base after solving problems, making design decisions, or discovering reusable patterns. Invoke whenever you have just fixed a bug, made an architectural choice, found a reusable pattern, learned something from external sources, or completed a refactoring. Capturing while context is fresh prevents knowledge loss across sessions.
Search the local knowledge base for past solutions, design decisions, and patterns before starting work. Invoke when facing an error or exception, when approaching an unfamiliar area of the codebase, or when considering implementation approaches. Prevents re-solving already-solved problems and surfaces relevant context proactively.
Get contextual knowledge suggestions based on the current work context — the file being edited, the task being performed, or the error being investigated. Invoke at the start of a work session, when opening a new file, or when beginning a feature. Surfaces relevant past knowledge proactively without requiring a specific query.
Analyze existing codebase to inform spec design. Scans specified packages or directories to extract architecture patterns, public APIs, dependency relationships, and test coverage. Output feeds directly into design.md.
| name | sdd-validate |
| description | Validate that implementation satisfies spec acceptance criteria. Maps each criterion in requirements.md to corresponding tests, runs them, and reports pass/fail per criterion. Use after implementation, before PR creation. |
Verify that the implementation actually satisfies what was specified. Catches gaps between requirements and implementation before the review phase.
$ARGUMENTS: Spec ID (e.g., KNOW-001) or path to spec directoryRead specs/KNOW-{NNN}-*/requirements.md and extract all acceptance criteria (AC-1, AC-2, ...).
For each acceptance criterion:
pnpm test:run
Compare implementation against design.md:
## Validation Report: KNOW-{NNN}
### Acceptance Criteria Coverage
| Criterion | Test File | Status |
| --------- | --------- | ----------------------- |
| AC-1: ... | tests/... | PASS / FAIL / UNCOVERED |
| AC-2: ... | tests/... | PASS / FAIL / UNCOVERED |
### Design Compliance
- [ ] Interfaces match design.md
- [ ] Data flow matches design.md
- [ ] Design decisions followed
### Build Status
pnpm verify: PASS / FAIL
### Coverage
pnpm test:coverage: XX% (target: 80%)
### Summary
X/Y acceptance criteria covered and passing.
Z criteria need attention.
docs/workflows/phase2-implementation.md — Implementation standards.ai-agents/config/review-checklist.md — Full review checklist