continuous-learning-v2

Transform Claude Code session observations into confidence-scored atomic instincts.

Updated Feb 2, 2026

Search Tags:#hooks #skills #learning #agents #local #evolve #instincts

Ready to run

npx skills add https://github.com/k3sdtw/.claude --skill continuous-learning-v2-k3sdtw

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: continuous-learning-v2
Source: https://github.com/k3sdtw/.claude/tree/main/skills/continuous-learning-v2
Command: npx skills add https://github.com/k3sdtw/.claude --skill continuous-learning-v2-k3sdtw

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Transform Claude Code sessions into reusable knowledge by turning observations into atomic instincts with confidence scoring, enabling continuous improvement without external data sharing.

Core Features & Use Cases

Observation hooks capture every tool usage event deterministically to surface actionable patterns.
Instincts are atomic, confidence-weighted units that evolve into commands, skills, and agents.
Local-first design: instincts and evolutions stay on-device and can be exported to teammates.

Quick Start

Turn your latest session observations into instincts and start evolving them into higher-level capabilities.

Frequently Asked Questions about continuous-learning-v2

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

FAQPage Schema

How do I transform Claude Code session observations into reusable instincts?▼

Observation hooks capture tool usage events from your sessions, converting them into atomic instincts with confidence scores. These instincts are stored locally in ~/.claude/homunculus and can be analyzed to identify patterns and evolve into higher-level commands, skills, and agents without external data sharing.

What's the best way to cluster session observations into actionable commands and skills?▼

Continuous analysis applies confidence-weighted instincts to cluster patterns across local workflows. The provided scripts generate, export, and evolve instincts deterministically, transforming raw observations into structured commands and skills that improve with each session.

Can I use instinct learning without sending data to external services?▼

Yes. The local-first design keeps all instincts and evolutions on-device by default. Your instincts repository stays under ~/.claude/homunculus, and you can optionally export instincts to teammates while maintaining full control over what leaves your machine.

Do I need external dependencies to generate and evolve instincts?▼

No external dependencies are required. The Skill uses local hooks and provided scripts to generate, export, and evolve instincts directly within your Claude Code environment, making it self-contained for continuous learning workflows.

How do instincts with confidence scores improve my Claude Code workflows over time?▼

Instincts are atomic, confidence-weighted units that capture actionable patterns from tool usage. As they accumulate and cluster across sessions, they surface reliable patterns that evolve into reusable commands, agents, and skills tailored to your specific workflows.