learning-extraction

You can extract actionable learnings from session transcripts by loading them into the extraction workflow, which applies an eight-category taxonomy, quality criteria, and confidence scoring to generate structured CLAUDE.md entries for future reuse.

The taxonomy used for categorizing extracted learnings is an eight-category classification system that structures insights from session transcripts, ensuring each extracted learning is sorted into a defined category before being added to your CLAUDE.md file.

Confidence scoring is applied automatically during the extraction workflow by evaluating extracted learnings against predefined quality criteria, ensuring only reliable insights with measurable confidence levels are surfaced for your CLAUDE.md integration.

Yes, you can process multiple session transcripts across different sessions. The extraction workflow identifies and structures actionable learnings across multiple transcripts, consolidating them into reusable CLAUDE.md entries with unified taxonomy and confidence scoring.

No, the learning extraction workflow does not modify your original session transcripts or user data. It reads reference materials and transcript patterns solely to produce new structured CLAUDE.md learnings without performing external actions or making changes.

The best way to document design session learnings for future improvements is to run the extraction workflow on your transcripts, which categorizes insights by taxonomy, applies quality criteria, and outputs structured CLAUDE.md entries to inform future projects.