reflect

You can auto-learn from corrections by running reflection analyses on your sessions. The skill identifies reusable patterns and proposes targeted updates to relevant skills to reduce recurring mistakes. It applies across multiple projects and sessions for consistency.

Cross-skill learning tracks correction patterns across different repos and sessions. When a lesson proves universal, it prompts promotion and syncs the updated patterns to the global configuration file to ensure consistent responses everywhere.

Yes, you can manually trigger a reflection to analyze a specific session. This process reviews the conversation history, identifies corrections and preferences, and proposes targeted updates to the relevant skills for your approval.

Yes, the skill update workflow requires manual approval before applying any changes. It uses a safe, advisory workflow to propose updates, ensuring you maintain control over modifications through manual approval and versioned backups.

The best way to safely back up skill updates is through the automatic reflection process. It runs at session end and automatically creates versioned backups of changes before syncing promoted lessons to the global configuration.

Claude Code may repeat mistakes without a cross-skill learning mechanism to track correction patterns. Analyzing sessions to identify reusable corrections and updating skills reduces recurring errors and improves consistency across projects.