AgentSkillsCN

postmortem

会话反思与反馈。回顾哪些环节效果良好、哪些环节尚有不足,并提出改进建议。可在会话结束时使用,或在被要求提供反馈时调用此技能。可通过 /postmortem 调用。

SKILL.md
--- frontmatter
name: postmortem
description: Session reflection and feedback. Reflects on what worked, what didn't, and suggests improvements. Use at end of session or when asked for feedback. Invoke with /postmortem.

Session Postmortem

Reflect on this session to help improve future agent prompts and workflows.

Answer These Questions

  1. Friction: What slowed you down or required extra effort?

  2. Unclear instructions: Were any parts of your prompt confusing or ambiguous?

  3. Missing context: What information did you need that wasn't provided?

  4. What worked well: What made this session smooth?

  5. Suggested changes: Specific improvements to prompts, workflows, or documentation?

  6. Subagents: Did you use Task() subagents? If so, what prompts worked well and what prompts produced poor results? Include the actual prompts.

  7. LSP plugins: Did the code intelligence plugins (rust-analyzer-lsp, pyright-lsp) help or distract? Did they catch real errors? Did they produce noise/false positives?

  8. Lint and config settings: Did any lint errors cause unnecessary churn (fixing things that didn't matter)? Were there warnings you wish had been enabled that would have caught real issues earlier? Suggest specific ruff/pyright/clippy rules to enable or disable.

Be concrete. "The prompt was unclear" is not useful. "The prompt said 'run tests' but didn't specify which command" is useful.