continuous-learning

To extract reusable patterns from Claude Code sessions, enable the stop hook to evaluate completed transcripts. The skill checks session length, detects error resolutions and workarounds, then saves these patterns into a persistent learned-skills directory for future reuse.

A stop hook for session pattern mining is an automation trigger that evaluates completed Claude Code conversations against configurable session-length thresholds. It identifies debugging techniques and workarounds, saving them as persistent skills after long troubleshooting sessions end.

Yes, Claude Code session analysis requires transcript-path access to evaluate conversations. The skill needs this access along with a shell-based evaluator and JSON config parsing to detect reusable debugging patterns and workarounds.

Yes, you can configure session-length thresholds for pattern extraction using JSON config parsing within the shell-based evaluator. This ensures only sufficiently long Claude Code sessions are reviewed for reusable debugging techniques and project-specific workarounds.

Debugging patterns extracted from Claude Code transcripts include error resolutions, user corrections, workarounds, and project-specific debugging techniques. These patterns are identified during session review and saved into a persistent learned-skills directory.