self-improve

To capture session learnings, use a session distillation process that extracts durable lessons from conversation context and routes them to project documentation like CLAUDE.md or AGENTS.md. This prevents valuable workarounds and corrections from being lost across development sessions.

Saving AI coaching insights requires intelligent routing that directs repeated guidance and skill-shaped knowledge to the auto memory layer. This ensures preferences and domain knowledge are continuously referenced and applied in future interactions.

Yes, you can extract corrections and failure modes from a debugging session and route them to existing skills or suggest new skills. This requires analyzing conversation context and project files to effectively filter and route skill-shaped knowledge.

You need session distillation when complex debugging uncovers specific workarounds, repeated guidance, or domain knowledge. Extracting these durable lessons ensures continuous improvement of AI models and prevents insights from being lost after the session ends.

Session distillation works without external dependencies or components, relying solely on analyzing conversation context, skill invocations, and project files. It routes extracted lessons to appropriate knowledge layers like project documentation, auto memory, or skills.

Session distillation extracts corrections, repeated guidance, skill-shaped knowledge, preferences, failure modes, domain knowledge, and improvement opportunities. These durable lessons are then routed to project documentation, auto memory, existing skills, or new skills.