AgentSkillsCN

Ralph Retro

拉尔夫复古

SKILL.md

/ralph-retro — Post-Session Retrospective

Usage

/ralph-retro

Description

Reviews the recent ralph loop iterations and extracts learnings. Updates AGENTS.md with new patterns, CLAUDE.md domain knowledge section, and identifies recurring issues.

Instructions

  1. Read PROGRESS.md — analyze the last session's iteration log entries.

  2. Read AGENTS.md — review existing learnings.

  3. Read recent git log (last 20 commits) for commit patterns and changes.

  4. Analyze for:

    • Recurring issues: Were the same types of problems hit multiple times?
    • Blocked tasks: Why were tasks blocked? Is there a pattern?
    • Task sizing: Were tasks the right size? Too big? Too small?
    • Missing conventions: Any patterns discovered that should be documented?
    • Gotchas: Any surprises that future iterations should know about?
  5. Present findings to the user as a brief retrospective report:

    • What went well
    • What didn't go well
    • Proposed additions to AGENTS.md
    • Proposed additions to CLAUDE.md domain knowledge section
  6. After user approval, update:

    • AGENTS.md with new learnings
    • CLAUDE.md domain knowledge section (between auto-generated markers)

Model

sonnet