菜单
Stress-test a plan/design for unknown unknowns using grounded blindspot lenses and one material decision at a time.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Drive one ready artifact through implementation, proof, review, and owner handoff.
Review the named UI scenario or emit one bounded improvement packet.
Teach a boring-ui child app how to enable Google signup with @hachej/boring-core. Use when the user asks for Google auth, Google OAuth, social signup, or how to turn on Google sign-in/sign-up in a child app.
Ask which Boring skill or workflow fits your situation. A router over the local Boring skills.
Run an extremely strict maintainability review for abstraction quality, giant files, spaghetti-condition growth, and code-judo simplification opportunities.
Route a request to the right Boring v2 workflow skill without doing the work.
| name | grill-for-unknowns |
| description | Stress-test a plan/design for unknown unknowns using grounded blindspot lenses and one material decision at a time. |
| disable-model-invocation | true |
Read the plan and real code/tests. Classify gaps before asking anything:
Probe seven lenses: scale, security, failure, edge cases, concurrency, migration, rollback. Rank by implementation risk.
Ask one question only when it is:
Use:
Question: <one decision>
Why: <consequences>
Evidence: <citation>
Recommendation: <default + rationale>
Resolve or explicitly accept every material unknown. Propose defaults for low-risk
gaps. Record hard-to-reverse decisions in docs/DECISIONS.md, not only chat.
Example: docs/issues/391/runtime-refactor/REVIEW-2026-07-11-unknowns.md.