continuous-learning-v2

The Skill transforms session observations into atomic instincts with confidence scoring, then clusters those instincts into commands, skills, and agents across your Claude Code workflows to create evolving knowledge artifacts.

To capture learning observations, you need to enable PreToolUse and PostToolUse hooks within your local Claude Code environment, which will feed session data into the observer pipeline for analysis and archiving.

Yes, you can use the /instinct-export command to export your learned instincts, allowing you to share insights, bootstrap team knowledge, and migrate structured observations across different environments.

The clustering process works by using a background observer agent to analyze and group atomic instincts based on their confidence scores, evolving them into organized capabilities like skills, commands, and agents in an evolved directory.

No external dependencies are required to run the continuous learning pipeline; it operates locally using scripts, requiring only the structured instincts catalog and the evolution pipeline to produce artifacts.

The best way to view generated instincts is by running the /instinct-status command, which displays your locally stored atomic instincts along with their confidence scores and current evolution state.