Session Retrospective

Priority: P1 (OPERATIONAL)

Structure

common/session-retrospective/
├── SKILL.md              # Protocol (this file)
└── references/
    └── methodology.md    # Signal tables, taxonomy, report template

Protocol

1. Extract — Scan for correction signals (loops, rejections, shape mismatches, lint rework)
2. Classify — Root cause: Skill Missing | Incomplete | Example Contradicts Rule | Workflow Gap | Trigger Miss
3. Trigger Miss Check — For every task in the session, ask: "Was a relevant skill available but not loaded?"
   • If yes: record skill ID, indirect phrase used, and fix (add keyword alias to triggers)
4. Propose — One fix per root cause: update skill, update reference, new skill, or new workflow
5. Implement — Apply to all agent dirs. Keep SKILL.md concise; move large tables to references/. Update AGENTS.md
6. Log to AGENTS_LEARNING.md — For each correction loop found, append one entry using common/common-learning-log protocol (Signal: Session retrospective)
7. Report — Output correction count, skills changed, trigger misses found, estimated rounds saved

Trigger Miss Output

Emit a trigger miss block (schema in references/methodology.md) for each miss detected.

Guidelines

• Cite specifics: Reference concrete conversation moment per proposal
• Extend first: Search AGENTS.md before creating — update existing skills
• One fix per loop: One correction → one targeted skill change
• Sync all agents: Apply to every agent skill dir listed in .skillsrc agents field
• Follow skill-creator: New skills comply with common/skill-creator standards

Anti-Patterns

• No Vague Proposals: Cite exact gap + fix, not "make X better"
• No Duplicate Skills: Search AGENTS.md index first
• No Oversized Patches: Extract to references/ per skill-creator standard

References

Signal tables, root cause taxonomy, report template, real-world example: references/methodology.md