Expand description
Post-mortem analysis: produce lessons and rule proposals from session data.
The analyzer takes session statistics and trigger reasons, then produces structured findings. The current implementation uses deterministic heuristics; future versions can delegate to LLM (Sonnet) for deeper analysis.
Output hierarchy (from GH-157 spec):
- Rules: Hook-enforced (block/auto-run), 100% compliance
- Lessons: CLAUDE.md auto-maintained paragraph, ~90% compliance
- Observations:
edda askon-demand, no enforcement
Structs§
- Analysis
Input - Session data available for analysis.
- Lesson
- A lesson extracted from post-mortem analysis (descriptive, not prescriptive).
- Post
Mortem Result - Complete result of a post-mortem analysis.
- Rule
Proposal - A rule proposal from post-mortem analysis (prescriptive).
Enums§
- Lesson
Severity - Severity of a lesson.
Functions§
- analyze
- Run deterministic post-mortem analysis on a triggered session.