菜单
Use when receiving code-review feedback, before implementing suggestions — especially when feedback seems unclear or technically questionable. Verify against the codebase and reason about each item instead of agreeing performatively or implementing blindly.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/dirge-code/dirge --skill code-review-feedback命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Use when starting a non-trivial feature or change — align on WHAT before HOW by recording a proposal, requirement deltas, and a task checklist with the `spec` tool, then archive to fold the change into the living specs.
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes — find the root cause first instead of guessing at symptom patches.
Use when creating, editing, or reviewing a dirge skill — how to structure SKILL.md, write a description that gets the skill loaded at the right time, keep it token-efficient, and verify it actually changes behavior.
基于 SOC 职业分类
正在显示 SKILL.md
| name | code-review-feedback |
| description | Use when receiving code-review feedback, before implementing suggestions — especially when feedback seems unclear or technically questionable. Verify against the codebase and reason about each item instead of agreeing performatively or implementing blindly. |
Code review is a technical evaluation, not an emotional performance.
Core principle: verify before implementing, ask before assuming, technical correctness over social comfort.
1. READ — the whole review, without reacting
2. UNDERSTAND — restate each item in your own words (or ask if unclear)
3. VERIFY — check it against codebase reality
4. EVALUATE — is it technically sound for THIS codebase?
5. RESPOND — technical acknowledgment, or reasoned pushback
6. IMPLEMENT — one item at a time, test each
Don't open with "You're absolutely right!", "Great point!", "Thanks for catching that!", or any gratitude/praise filler. State the fix instead, or just make it — the code shows you heard the feedback.
Fixed — <what changed> in <where>.If any item is unclear, stop — don't implement the clear ones yet. Items are
often related, and partial understanding produces the wrong implementation. Ask
for clarification (the question tool), presenting your interpretations as
options.
"I understand items 1, 2, 3, and 6. I need clarification on 4 and 5 before implementing."
Feedback from a human reviewer or an automated reviewer is a suggestion to evaluate, not an order to follow. Before implementing, check:
If a suggestion seems wrong, push back with technical reasoning — reference the tests/code that show it. If you can't verify it, say so: "I can't verify this without X — should I investigate, ask, or proceed?" If it conflicts with a prior decision the user made, stop and raise that first.
If a reviewer asks you to "implement this properly," grep for actual usage first. If nothing calls it: "This isn't called anywhere — remove it (YAGNI)?" If it is used, then implement it properly.
State it factually and move on — no long apology, no defending why you pushed back:
"You were right — I checked X and it does Y. Implementing now."
When replying to inline review comments, reply in the comment thread
(gh api repos/{owner}/{repo}/pulls/{pr}/comments/{id}/replies), not as a
top-level PR comment.
Verify. Question. Then implement. Technical rigor over performative agreement.
Adapted from superpowers by Jesse Vincent (MIT License).