AgentSkillsCN

self-review

根据任务复杂度灵活调整评审力度。

SKILL.md
--- frontmatter
name: self-review
description: Scale review effort to task complexity
user-invocable: false
model: sonnet

Adaptive Review

Scale your effort to the task. Don't over-review trivial changes, don't under-review critical ones.

Effort Levels

Task TypeReview Depth
Typo, one-linerDoes it work? Ship it.
Feature, componentBuild + types + looks right
Architecture, refactorAll above + system impact + docs

Quick Check

After any change:

  1. •npm run typecheck && npm run build - must pass
  2. •Does it solve the problem? (not just technically correct)
  3. •Would I approve this PR?

If yes to all, move on.

When to Dig Deeper

More review: Money, security, user data, unfamiliar code Less review: Isolated changes, low risk, well-understood code

→ For detailed verification workflow, use verify command.