Agent Skills › WingedGuardian/GENesis-AGI › retrospective

retrospective

GitHub

用于在任务或交互完成后进行回顾分析，提取经验教训、识别流程改进点并更新程序记忆。适用于多步骤任务结束、发现意外情况或解决障碍后，通过重构时间线和总结模式来优化后续工作。

src/genesis/skills/retrospective/SKILL.md WingedGuardian/GENesis-AGI

Trigger Scenarios

多步骤任务完成（成功或失败）用户交互中出现意外或差距障碍被解决后主动工作会话结束时

Install

npx skills add WingedGuardian/GENesis-AGI --skill retrospective -g -y

More Options

Non-standard path

npx skills add https://github.com/WingedGuardian/GENesis-AGI/tree/main/src/genesis/skills/retrospective -g -y

Use without installing

npx skills use WingedGuardian/GENesis-AGI@retrospective

指定 Agent (Claude Code)

npx skills add WingedGuardian/GENesis-AGI --skill retrospective -a claude-code -g -y

安装 repo 全部 skill

npx skills add WingedGuardian/GENesis-AGI --all -g -y

预览 repo 内 skill

npx skills add WingedGuardian/GENesis-AGI --list

SKILL.md

Frontmatter

{
    "name": "retrospective",
    "phase": 6,
    "consumer": "cc_background_reflection",
    "skill_type": "workflow",
    "description": "Post-interaction retrospective analysis — use after completing a significant task, conversation, or phase to extract lessons, identify process improvements, and update procedures"
}

Retrospective

Purpose

Analyze a completed interaction or task to extract lessons, identify process improvements, and update procedural memory.

When to Use

After a multi-step task completes (success or failure).
After a user interaction that revealed a gap or surprise.
After an obstacle was resolved (capture the resolution pattern).
Scheduled: end of each active work session.

Workflow

Reconstruct timeline — What happened, in what order? What was the goal?
Identify outcomes — Did it succeed? Partially? What was the quality?
Extract surprises — What was unexpected? What assumptions broke?
Find patterns — Does this match any existing procedural knowledge? Does it contradict any?
Derive lessons — Concrete, actionable learnings (not vague platitudes).
Update memory — Write observations, update procedures if warranted, flag contradictions for user review.

Output Format

subject: <what was analyzed>
date: <YYYY-MM-DD>
outcome: success | partial | failure
surprises:
  - <unexpected finding>
lessons:
  - <concrete actionable lesson>
procedure_updates:
  - procedure: <name>
    change: <what to update>
observations:
  - <observation to store>