AgentSkillsCN

prioritize-review-comments

在接收来自钩子、GitHub 机器人或评审代理的反馈后,决定哪些自动化评审意见需要修复,哪些可以跳过。

SKILL.md
--- frontmatter
name: prioritize-review-comments
description: Decide which automated review comments to fix vs skip. Use after receiving feedback from hooks, GitHub bots, or review agents.

Prioritize Review Comments

After receiving automated review feedback, apply this framework to decide what to fix.

What to Fix

Fix these regardless of severity labels:

  • Issues directly related to the code you changed
  • Problems in the feature you're implementing
  • Anything that makes the current PR less clean
  • "Minor" issues that take seconds to fix
  • "Deferrable" issues that are actually in-scope

The label doesn't matter. If it's in the scope of your change and it's a valid point, fix it.

What NOT to Fix

Don't scope-creep:

  • Unrelated code that happens to be nearby
  • Pre-existing issues in files you touched
  • Suggestions that turn a small PR into a large refactor
  • "While you're here..." improvements to other features

Don't fix incorrect feedback:

  • Comments where the reviewer misunderstood the change
  • Suggestions based on wrong assumptions
  • "Improvements" that would break intended behavior

Decision Framework

For each review comment, ask:

  1. Is it correct? If the reviewer misunderstood, skip it.
  2. Is it in scope? If it's about the feature/change you're making, fix it.
  3. Does the severity label matter? No. "Minor" in-scope issues should still be fixed.

Example Responses

Fix: "Minor: variable name could be clearer" → If it's in your changed code, rename it.

Fix: "Tiny nit: missing newline at end of file" → Takes 2 seconds, do it.

Skip: "While you're here, the function above could use better error handling" → Out of scope, different change.

Skip: "This should use the new API pattern" → Out of scope if you're intentionally matching existing code.