continuous-learning-v2

Extract atomic instincts from Claude Code sessions into local knowledge.

Updated May 1, 2026
One-click install
npx skills add https://github.com/oguzhanguvenkaya/oguzhan_claude_code_configurations --skill continuous-learning-v2-oguzhanguvenkaya
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: continuous-learning-v2
Source: https://github.com/oguzhanguvenkaya/oguzhan_claude_code_configurations/tree/main/skills/continuous-learning-v2
Command: npx skills add https://github.com/oguzhanguvenkaya/oguzhan_claude_code_configurations --skill continuous-learning-v2-oguzhanguvenkaya

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Turn Claude Code sessions into reusable knowledge by extracting atomic instincts with confidence scoring, enabling long-term skill evolution.

Core Features & Use Cases

  • Atomic "instincts" with confidence weighting that capture small learned behaviors from session observations.
  • Hooks-based observation (PreToolUse/PostToolUse) that reliably capture prompts and tool usage, feeding a background agent for pattern detection.
  • Evolution path from instincts to evolving into commands, skills, or agents, with local storage and export/import capabilities.
  • Configurable privacy and local storage, with a self-contained workflow that builds an evolving knowledge base without sending data externally.

Quick Start

Enable observation hooks, initialize the local homunculus structure, and begin capturing Claude Code sessions to start building instincts.

Frequently Asked Questions about continuous-learning-v2

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

FAQPage Schema
How do I capture Claude Code sessions to build a reusable knowledge base?

You capture Claude Code sessions by enabling PreToolUse and PostToolUse hooks that observe prompts and tool usage. The observation hooks feed a background agent that detects patterns and extracts atomic instincts into a local homunculus directory.

What are atomic instincts in a code learning system?

Atomic instincts are small learned behaviors extracted from session observations with confidence weighting. They capture reusable knowledge from your coding sessions and evolve into commands, skills, or agents for long-term learning automation.

Does this code learning system send my session data externally?

No, the code learning system stores all data locally under ~/.claude/homunculus. It provides configurable privacy and a self-contained workflow that builds an evolving knowledge base without sending data externally.

How do I turn extracted instincts into reusable commands or agents?

You turn extracted instincts into reusable commands or agents through a local evolution workflow. The system provides an evolution path that transforms atomic instincts with confidence scoring into commands, skills, or agents, with export and import capabilities.

Do I need any specific dependencies to start tracking session observations?

No specific dependencies are required to start tracking session observations. You need to enable observation hooks, initialize the local homunculus structure, and provide a YAML frontmatter containing name and description to begin capturing sessions.

Why use hook-based observation for extracting learned coding behaviors?

Hook-based observation reliably captures prompts and tool usage during live Claude Code sessions. Using PreToolUse and PostToolUse hooks feeds a background agent for pattern detection, ensuring accurate extraction of atomic instincts with confidence scoring.