continuous-learning

You can automatically extract reusable patterns from Claude Code sessions by enabling a stop hook that evaluates session transcripts at the end of each run and saves detected patterns as learned skills.

Continuous learning in Claude Code is an automated pattern extraction process that evaluates session transcripts against a minimum context threshold at session end to identify debugging techniques and workarounds for future reuse.

Yes, you can customize where learned skills are saved by modifying the config.json file, though by default the system persists extracted patterns directly into the ~/.claude/skills/learned/ directory.

To start auto-learning project conventions, install and enable the stop hook, then run Claude Code sessions as usual so the skill can automatically evaluate transcripts and store actionable patterns.

Yes, the pattern extraction mechanism evaluates transcripts against a minimum context threshold, ensuring only sessions with sufficient interaction depth are processed to identify actionable patterns.

Claude Code can automatically detect actionable patterns such as debugging techniques, workarounds, and project-specific conventions from your session transcripts and save them for future use.