一键导入
review-code
Code review workflow. Delegates to the QA agent to review code changes, a pull request, or recent commits. Use when asked to review code or a PR.
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。
菜单
Code review workflow. Delegates to the QA agent to review code changes, a pull request, or recent commits. Use when asked to review code or a PR.
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。
Systematic debugging workflow. Analyzes error context, maps relevant code, forms and tests hypotheses, then implements and verifies a fix with team lead oversight. Use when investigating bugs or unexpected behavior.
End-to-end feature implementation workflow. Creates a todo with description and a full implementation plan, implements with the appropriate developers, and runs QA review with team lead oversight. Use when asked to implement a new feature or make substantial changes.
基于 SOC 职业分类
| name | review-code |
| description | Code review workflow. Delegates to the QA agent to review code changes, a pull request, or recent commits. Use when asked to review code or a PR. |
Identify what to review:
git diffgit diff main...<branch>Or use the /review command for interactive mode selection.
Use the team tool to delegate to the qa agent. The QA agent has access to report_finding and submit_review tools for structured output.
For uncommitted changes:
team run qa "Perform a structured code review of uncommitted changes. Run `git diff && git diff --staged` to see changes. For each issue, use `report_finding` with priority (P0-P3), confidence, file path, and line range. After all findings, call `submit_review` with verdict and summary."
For branch reviews:
team run qa "Perform a structured code review of branch <branch> compared to main. Run `git diff main...<branch>`. For each issue, use `report_finding` with priority (P0-P3), confidence, file path, and line range. After all findings, call `submit_review` with verdict and summary."
For commit reviews:
team run qa "Review changes in commit <hash>. Run `git show <hash>`. For each issue, use `report_finding`. After all findings, call `submit_review`."
The review produces structured findings with priorities:
Verdicts:
If the QA agent identifies blocking issues (P0/P1), delegate fixes to the appropriate developer:
team run dev-backend "<fix description based on QA findings>"
Then re-run QA to verify the fixes.