continuous-learning-v2

You can automate learning coding patterns by capturing tool-use observations during development sessions to generate atomic, confidence-weighted instincts. This Skill analyzes your tool use to distill repetitive corrections into reusable, project-scoped behaviors automatically.

Keep project-specific framework conventions isolated by operating within project-scoped directories. This approach isolates framework-specific patterns, like React conventions, while promoting universal best practices to a global scope for broader reuse.

Claude analyzes development sessions using background Haiku-based analysis to cluster related observations. As your knowledge base grows, this evolution pipeline groups atomic instincts into full-fledged reusable commands and agents.

Check confidence levels of learned behaviors by running the instinct status command. This displays all currently captured patterns and their assigned confidence scores for your active project environment.

Yes, you can stop repetitive manual corrections by automatically observing your sessions and distilling them into reusable instincts. It assigns confidence scores to extracted patterns, eliminating inconsistent coding behaviors over time.

No external dependencies are required to extract instincts. The Skill operates independently using background Haiku-based analysis to capture tool-use patterns and cluster them into reusable commands without additional packages.