continuous-learning

To extract reusable patterns from Claude Code sessions, a stop-hook workflow analyzes session work at the end of each run. It identifies patterns like error_resolution and debugging_techniques, then saves them to ~/.claude/skills/learned/ for future reuse.

A stop hook is an automated trigger that evaluates your Claude Code session at the end of each run. It detects user_corrections and workarounds, minimizing overhead by running only when the session concludes rather than continuously monitoring.

Yes, you can configure thresholds and paths for learned skill extraction using the config.json file. This configuration allows you to control how patterns are detected and where extracted skills are stored within your Claude Code environment.

The pattern detection mechanism identifies error_resolution, user_corrections, workarounds, and debugging_techniques. These categorized patterns are extracted from your Claude Code sessions and saved as learnable skills for quick reuse across projects.

Yes, learned skills apply across Claude Code sessions and projects. Extracted patterns are persisted to the user's home directory under ~/.claude/skills/learned/, ensuring reusable knowledge is available regardless of which project you are working on.

No external dependencies are required to automate end-of-session analysis. The skill operates independently using the stop-hook workflow and saves extracted patterns directly to your local filesystem without needing additional packages or libraries.