What is an 'instinct' in this system?

An instinct is a small, atomic learned behavior—such as a preference for functional programming—that includes a trigger, an action, and a confidence score based on evidence.

Is my private code shared or uploaded during the learning process?

No. All observation data and learned instincts are stored locally on your machine. Data is only shared if you manually use the export command to share your patterns.

How does Continuous Learning v2 differ from the previous version?

Version 2 moves from probabilistic skill-based observation to 100% reliable deterministic hooks, and introduces an 'instinct' model with confidence scoring for better accuracy.

How do I convert learned patterns into permanent tools?

By using the /evolve command, the system clusters related instincts together and generates structured SKILL.md files, commands, or specialist agents automatically.

Continuous Learning v2

Name: Continuous Learning v2
Author: xu-xiang

byxu-xiang

•

323

•

Productivity & Workflow

Transforms Claude Code sessions into reusable knowledge by observing patterns and evolving them into specialized skills and agents.

Continuous Learning v2 is an advanced behavioral adaptation system for Claude Code that utilizes deterministic hooks to capture user patterns, corrections, and workflow preferences. By breaking down complex interactions into atomic 'instincts' with confidence scoring, the system creates a personalized knowledge base that evolves from simple observations into high-level commands, skills, and autonomous agents. This ensures Claude grows more efficient with every interaction, specifically tailoring its responses to your unique coding style and project requirements.

Key Features

01Deterministic observation via PreToolUse and PostToolUse hooks for 100% reliability

02Atomic instinct generation with dynamic confidence weighting from 0.3 to 0.9

03Full export and import capabilities for sharing learned behaviors across machines

04323 GitHub stars

05Evolutionary clustering that transforms repeated patterns into full skills and commands

06Background pattern detection using efficient AI models for minimal latency

Use Cases

01Creating specialized AI agents that understand specific debugging or testing methodologies

02Automatically capturing and enforcing project-specific coding styles and preferences

03Building a library of custom commands based on repeated manual terminal workflows

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add xu-xiang/everything-claude-code-zh continuous-learning-v2

For use in Claude.ai and ChatGPT

Key Features

01Deterministic observation via PreToolUse and PostToolUse hooks for 100% reliability

02Atomic instinct generation with dynamic confidence weighting from 0.3 to 0.9

03Full export and import capabilities for sharing learned behaviors across machines

04323 GitHub stars

05Evolutionary clustering that transforms repeated patterns into full skills and commands

06Background pattern detection using efficient AI models for minimal latency

Use Cases

01Creating specialized AI agents that understand specific debugging or testing methodologies

02Automatically capturing and enforcing project-specific coding styles and preferences

03Building a library of custom commands based on repeated manual terminal workflows

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add xu-xiang/everything-claude-code-zh continuous-learning-v2

For use in Claude.ai and ChatGPT