continuous-learning-v2

Continuous learning from CLI tool-use sessions is automated by capturing PreToolUse and PostToolUse hooks, converting recurring tool-use patterns into reusable, confidence-scored instincts. This reduces repeated work by evolving your tool-preference routines and error-resolution habits over time.

To extract and reuse recurring tool-use patterns, the system observes deterministic tool hooks and persists data locally in JSONL format. It clusters related patterns into atomic instincts with evidence and confidence scores from 0.3 to 0.9.

Yes, you can share learned workflow automation patterns between machines. The system supports import and export functionalities for moving learned instincts between users without sharing raw conversation content.

Learned instincts evolve into broader workflow automation skills through clustering related trigger-driven instincts. This process prepares structured outputs that can evolve into standalone skills, commands, and agents for repeated development tasks.

Yes, instinct extraction requires deterministic hook-based observation to capture CLI tool activity. You must configure PreToolUse and PostToolUse hooks to ensure recurring behaviors and tool-preference routines are not missed during sessions.

The confidence scoring range used for tracking learned instincts in workflow automation is 0.3 to 0.9. This scoring includes evidence tracking to validate trigger-driven atomic instincts captured during your development and automation tasks.