菜单
Use when you need to ask the user clarifying questions, gather requirements, make a decision that requires user input, or confirm anything before proceeding with a task.
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
Use when implementing background jobs, task queues, scheduled jobs, retries, or any Redis-backed async worker in a Python backend service in this harness. arq is the chosen queue (async-native, Redis-backed). Replaces the old bullmq skill.
Use when creating a new branch, starting new feature work, fixing a bug on a new branch, or any situation where a branch needs to be created. Use when the user asks to "create a branch", "start a new branch", "branch off of X", or "make a branch for this work".
Use when creating CI/CD pipelines for any project. Use when asked to "set up CI/CD", "create a pipeline", "automate deployments", "configure GitHub Actions", "set up Vercel/Render/AWS", or similar. The harness is a polyglot monorepo — a Python (FastAPI + uv) backend in `apps/api/` and a TypeScript (Vite + pnpm) frontend in `apps/web/`. Pipelines run the two stacks as parallel lanes that join at a final gate.
Use when writing, reviewing, or editing a Dockerfile or docker-compose file. Use when asked to "containerize", "dockerize", "add Docker support", or "write a Dockerfile" for any application or service in this harness — FastAPI backend (Python + uv) or Vite frontend (TypeScript + pnpm, served as static assets).
Use when writing, reviewing, or refactoring TypeScript code on the frontend — especially when tempted to use `any`, type assertions, unvalidated casts, or when designing types, generics, utility types, or tsconfig settings. Backend code in this harness is Python, not TypeScript; see `effective-python` / `python-linting` for the backend equivalents.
FastAPI implementation guide. Load whenever building or modifying any backend HTTP service in this harness — defining routes, request/response models, dependencies, middleware, error handlers, lifespan events, or background tasks. Triggers include: APIRouter, Depends, FastAPI app instance, response_model, OpenAPI generation, async endpoint handlers. Use alongside `sqlalchemy`, `pydantic`, and `rest-api-design`.
| name | plain-text-questions |
| description | Use when you need to ask the user clarifying questions, gather requirements, make a decision that requires user input, or confirm anything before proceeding with a task. |
Ask questions as a numbered list in plain text. The user responds inline. Never use the question tool.
Never use the question tool. Write questions directly in your message as a numbered list, then stop and wait for the user to reply before doing anything else.
Before I proceed, I have a few questions:
1. What framework is this project using?
2. Should this replace the existing implementation or live alongside it?
3. Are there specific components you want to prioritize first?
Stop after asking. Do not start work, explore the codebase, or take any action until the user replies.
question tool — never do this, regardless of how simple the question seems