learning-taxonomy

Classify learnings into six categories—caveat, pattern, error_fix, dependency, command, architecture—then route them to appropriate memory files. This guides targeted improvements to your project knowledge and Claude Code context, making future sessions more effective.

Learning taxonomy includes caveat (constraints discovered), pattern (repeatable approaches), error_fix (solutions to problems), dependency (tool or library requirements), command (CLI or operational steps), and architecture (structural decisions). Each category routes to specific memory locations for project-wide or file-specific updates.

Map learnings to project-wide, topic-specific, or file-specific memory routing rules defined in CLAUDE.md. The Skill provides routing paths and templates to update your memory structure, ensuring knowledge persists and informs future work.

Yes. After a session, classify key takeaways—such as environment setup requirements or discovered patterns—into the six learning categories, then use routing guidance to update your project memory. This captures actionable improvements for follow-up work.

A caveat captures constraints or preconditions you discovered (e.g., 'environment setup required before follow-up'); a pattern documents a repeatable approach or technique. Routing them separately ensures caveats inform prerequisites while patterns guide methodology.