Continuous Learning v2 is an advanced framework designed to make Claude Code smarter with every interaction. By utilizing PreToolUse and PostToolUse hooks, it captures 100% of session activity to identify 'atomic instincts'—small, weighted behavioral patterns such as specific coding styles, error resolutions, and repeated workflows. These instincts are processed by a background observer agent, assigned confidence scores, and eventually evolved into full-fledged custom skills, commands, or specialized agents, creating a truly adaptive AI assistant tailored to your unique development environment.
主要功能
01Deterministic observation via system hooks for 100% reliable data capture
02Privacy-first local storage with granular export/import capabilities for sharing
03Automated evolution of learned patterns into reusable skills and commands
0461 GitHub stars
05Background pattern detection using efficient Haiku-based observer agents
06Atomic instinct modeling with confidence-weighted scoring (0.3 to 0.9)
