菜单
Repo-local overlay of the cross-family grill-me skill. Stress-test a plan or design through one-question-at-a-time interviews, with a 6-item completion checklist required before Plan Review for medium+ or sensitive tasks.
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
| name | grill-me |
| description | Repo-local overlay of the cross-family grill-me skill. Stress-test a plan or design through one-question-at-a-time interviews, with a 6-item completion checklist required before Plan Review for medium+ or sensitive tasks. |
Interview the user (or the primary agent's draft plan) relentlessly about every aspect of the design until reaching shared understanding. Walk down each branch of the decision tree, resolving dependencies one at a time. For each question, provide your recommended answer and its reasoning.
Ask questions one at a time. If a question can be answered by exploring the codebase, explore the codebase instead of asking.
When any of the following holds, the griller's AI family must differ from
the primary agent's family (see .agents/reviewers/routing-matrix.md for the
three-family definition):
{medium, large, huge}Refactoring risk in {medium, high}small and tiny tasks may use same-family or skip grill-me entirely.
When the griller writes its result into the review-gate payload, set
mode: "grill" and include primaryFamily, grillerFamily, and taskSize.
Older payloads without mode are treated as mode: "review" for backwards
compatibility.
Before the plan may proceed to the Plan Review checkpoint, the grill summary must record concrete answers (not "N/A") for all six items below. If any high-impact item remains unresolved, do not exit grill mode.
confirmed / changed / dropped.sync-skills.ps1 propagation is needed; if yes, list the sibling
repos and files in scope.high-impact or low-impact. Any high-impact unresolved item
blocks exit.When grilling concludes, emit:
Grill summary
- Primary family: <copilot|gemini|codex>
- Griller family: <copilot|gemini|codex> (must differ for medium+/sensitive)
- Task size: <tiny|small|medium|large|huge>
- Refactoring risk: <none|low|medium|high>
- Sensitive surface: <yes|no>
Checklist
1. Assumptions resolved: ...
2. Failure modes: ...
3. Rollback path: ...
4. Test surface: ...
5. Sibling-repo impact: ...
6. Open questions remaining: <none high-impact | listed below>
Verdict
- Ready for Plan Review: yes/no
- Reason: ...
The reviewer at the Plan Review checkpoint will refuse to sign if the grill
summary is missing or incomplete (see must_refuse_when in
.agents/reviewers/common-review-contract.toml).
Pre-flight checklist that every primary agent runs before opening the pre-implementation gate. Confirms task size, primary AI family, cross-family reviewer routing, grill-me requirement, and the exact `pnpm review:approve-pre-implementation` flags to use.
Discovers and invokes the repo-local skill set. Use when starting work or when you need to choose the right workflow for the current task.
Repo-specific Angular conventions and patterns for law-prep-web. Load when working with Angular components, services, DI, signals, routing, or tests.
Delivers changes incrementally. Use when implementing any feature or change that touches more than one file. Use when you're about to write a large amount of code at once, or when a task feels too big to land in one step.
Import, merge, or combine repositories into an Nx workspace using nx import. USE WHEN the user asks to adopt Nx across repos, move projects into a monorepo, or bring code/history from another repository.
Breaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible.