Black-Tortoise Agent Quality Scan Skill
When to use
- Before and after a PR, when you want to quickly list high-risk points (lint/type/test failures, API misuse, layer violations).
- When you need to reduce the scope of changes and produce an actionable fix list.
Workflow
- Scope locking: use
search/changes or git diff to determine the files to scan; if there is no diff, focus on the current feature directory.
- Rule location: according to
docs/INDEX.md , read the corresponding AGENTS.md / README.md, and confirm layer boundaries and prohibited items.
- Issue collection：
- Enable
codacy/* to obtain analysis; record critical/high-priority items.
- If there are framework API concerns, use
context7/* to check the official signatures.
- Breakdown and prioritization: use
sequentialthinking/* to produce minimal fix steps (high risk, low cost first).
- Suggested fixes: for each issue, provide the minimal actionable fix and file path; avoid cross-layer changes.
- Verification: recommend running
pnpm run lint、pnpm run architecture:gate, and related tests (do not run automatically unless requested).
Output format
- Assumptions/risks → issue list (with file links) → suggested fix steps → suggested verification.
- Mark items that require user decisions (such as changing APIs or adjusting configuration).
Prohibited items
- Over-fixing at the cost of increasing scope; keep changes minimal.
- Introducing RxJS state management or
as any。
- breaking the one-way layer dependency or deep imports across capabilities.
