AgentSkillsCN

retrospective

在完成一项重要任务后运行,回顾哪些环节进展顺利,哪些方面仍有改进空间。

SKILL.md
--- frontmatter
name: retrospective
description: Run after completing a significant task to reflect on what went well and what could improve.
disable-model-invocation: true

After completing a significant task, run a brief retrospective:

  1. Read the conversation history -- What was asked? What was delivered?
  2. Evaluate:
    • Did the solution stay within scope?
    • Were changes small and reversible?
    • Were there any wasted steps or wrong turns?
    • Was anything harder than expected? Why?
  3. Extract insights -- If there's a reusable lesson, append it to .claude/lessons.md.
  4. Report to the user -- Share a 2-3 sentence summary of what went well and one thing to improve next time.

Keep it honest and brief. The goal is continuous improvement, not ceremony.