session-learner

Session-learner analyzes your entire conversation to extract learnings across user preferences, agent failure patterns, technical patterns, and workflows. Say 'learn from this session' to identify and categorize insights automatically, then persist them to designated destinations like CLAUDE.md files or project documentation.

Session analysis identifies five categories of learnings: user preferences in interaction style, agent failure patterns revealing process gaps, project technical patterns for consistency, reusable workflows for future efficiency, and reference knowledge for documentation. Each is categorized and routed to the appropriate storage location.

Session-learner detects conflicts when new learnings overlap with existing CLAUDE.md or project documentation. It surfaces conflicts explicitly, proposes merge or replacement options, and presents findings for your review before updating any files, preventing accidental overwrites.

Yes. Session-learner builds a progressive documentation record by persisting categorized learnings to user globals and project docs. Each session adds to the structured knowledge base, enabling pattern recognition and better collaboration as learnings accumulate.

Session-learner requires no external dependencies. It analyzes your current conversation and writes to standard destinations (CLAUDE.md files or project docs/). No configuration or prerequisites are needed to start capturing insights.