菜单
Use when assigning work to a worker or subagent that cannot see the main conversation, especially when the prompt must be self-contained, precise, and resistant to ambiguity or lazy follow-up delegation
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 verifying implementation work and the failure mode is superficial approval, code-reading in place of execution, or over-trusting a passing test suite without trying to break the change
Use when coordinating multi-step work across workers or subagents, especially when research, implementation, and verification should be split while the main agent stays responsible for synthesis and user communication
Use when the task is codebase exploration rather than modification, and you need a fast, read-only search pattern that starts broad when location is unknown, narrows when location is known, and parallelizes independent searches
Use when deciding what information belongs in memory, how to classify it, and what must not be saved because it is derivable from the repo, git history, or current task context
Use when deciding whether shell commands are truly read-only, safe for concurrent execution, and free of path, git, redirection, or command-shape risks that should force a stricter permission path
Use when the real risk is not lack of knowledge but self-justifying shortcuts such as explaining instead of running commands, skipping the hard part first, or deciding on your own that verification would take too long
| name | worker-prompt-craft |
| description | Use when assigning work to a worker or subagent that cannot see the main conversation, especially when the prompt must be self-contained, precise, and resistant to ambiguity or lazy follow-up delegation |
Workers do not share your conversational context. A good worker prompt must stand on its own.
This skill follows the explicit prompt rules in src/coordinator/coordinatorMode.ts.
Never assume the worker knows what you know from the user conversation.
Forbidden prompt shapes:
Those prompts outsource context reconstruction to the worker and usually produce drift.
Every worker prompt should include:
Open with why this task exists:
The purpose helps the worker choose the right depth.
Ask for:
Do not ask for edits in the same prompt unless that is truly the task.
Ask for:
Ask for:
Use a fresh verifier when independence matters.
Continue the same worker when its existing context is an asset.
Spawn a fresh worker when:
Fix the null check in src/auth/validate.ts:42. Session.user can be undefined when the session is expired but the token is still cached. Return 401 with "Session expired" instead of dereferencing user.id. Update the relevant tests in src/auth/validate.test.ts. Run the targeted tests and report the commit hash.
You have applied this skill correctly when a fresh worker could complete the task from your prompt alone without seeing any of the surrounding conversation.
