self-improvement

To enable agent self-improvement, use a systematic skill update workflow that analyzes past mistakes, identifies root causes, and modifies existing skills to prevent future failures.

The best way to document a mistake is triggering an error reflection process that generalizes the error pattern and updates the agent's knowledge base with new preventative rules to avoid repeating the error.

You should update an existing skill rather than creating a new one when capturing a learning pattern, always searching existing skills first and creating new skills only when the update is fully justified.

To sync local skills with upstream changes from an AI development kit, execute an upstream source sync workflow that tracks and integrates updates from the databricks-solutions/ai-dev-kit repository.

You should trigger a self-reflection workflow after encountering errors, completing complex tasks, or when asked to reflect, learn, document a mistake, or prevent a specific error pattern from recurring.