continuous-learning-v2

Extract reusable knowledge patterns from Claude Code sessions via hooks.

2|Updated Mar 12, 2026
One-click install
npx skills add https://github.com/sayasaya8039/ZWG_Terminal --skill continuous-learning-v2-sayasaya8039
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: continuous-learning-v2
Source: https://github.com/sayasaya8039/ZWG_Terminal/tree/main/.claude/skills/continuous-learning-v2
Command: npx skills add https://github.com/sayasaya8039/ZWG_Terminal --skill continuous-learning-v2-sayasaya8039

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Claude Code sessions contain valuable patterns, corrections, and workflows that disappear after the session ends. This Skill automatically captures these moments through hooks, converts them into atomic "instincts" with confidence scores, and evolves them into reusable skills, commands, or agents. It prevents knowledge loss and helps you build a personalized library of best practices that improve over time.

Core Features & Use Cases

  • Automatic Session Observation: Hooks capture every tool use and user correction in real-time, ensuring no pattern is missed.
  • Project-Scoped Learning: v2.1 isolates instincts per project so React patterns stay in React projects and Python conventions stay in Python projects, preventing cross-project contamination.
  • Confidence-Weighted Evolution: Instincts start with confidence scores (0.3-0.9) based on observation frequency and can be promoted to global scope when they appear across multiple projects.
  • Use Case: A developer working on multiple codebases can automatically learn project-specific conventions like "use functional components" for one project and "prefer dataclasses" for another, while universal patterns like "always validate input" are shared globally.

Quick Start

Add the provided hook configuration to your Claude Code settings, open any project directory, and begin your normal workflow to let the system automatically capture and learn your coding patterns.

Frequently Asked Questions about continuous-learning-v2

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automatically save coding patterns from Claude Code sessions?

You can automatically save coding patterns by registering PreToolUse and PostToolUse hooks that observe your Claude Code sessions, extract reusable knowledge, and convert it into atomic instincts with confidence scores.

What is project-scoped learning for coding conventions?

Project-scoped learning isolates captured instincts per project using git remote URL hashing, ensuring React patterns stay in React projects and Python conventions remain in Python projects without cross-contamination.

How do I prevent knowledge loss of workflow corrections in Claude Code?

To prevent knowledge loss, a background observer agent captures every tool use and user correction in real-time through hooks, evolving observed behaviors into reusable skills, commands, or agents.

Can I use this pattern detection system across multiple codebases?

Yes, the system supports multiple codebases by applying project-scoped learning to capture project-specific conventions locally, while promoting universal patterns to global scope when they appear across multiple projects.

How do confidence scores work for retained coding instincts?

Confidence scores range from 0.3 to 0.9 based on observation frequency, determining how strongly an instinct is retained before it evolves into a reusable skill or command.

Do I need to configure hooks to start continuous learning in Claude Code?

Yes, you need to add the provided hook configuration to your Claude Code settings to enable the background observer agent that captures your coding patterns during normal workflows.