continuous-learning-v2

Translate Claude Code session data into YAML instinct artifacts with confidence scores.

2|Updated Mar 20, 2026
One-click install
npx skills add https://github.com/multiplex-ai/muggle-ai-teams --skill continuous-learning-v2-multiplex-ai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: continuous-learning-v2
Source: https://github.com/multiplex-ai/muggle-ai-teams/tree/main/skills/continuous-learning-v2
Command: npx skills add https://github.com/multiplex-ai/muggle-ai-teams --skill continuous-learning-v2-multiplex-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Instinct-based learning turns Claude Code sessions into reusable knowledge by converting observed behaviors into atomic instincts with confidence scoring, which can later evolve into skills, commands, or agents in a robust workflow.

Core Features & Use Cases

  • Hooks-based observation captures tool usage during Claude Code sessions and stores them as instinct prompts with provenance.
  • Atomic instincts with confidence scoring enable measurable evolution into higher-level skills, commands, or agents.
  • Project-scoped instincts isolate patterns per repository, while a global layer supports universal patterns; cross-project contamination is avoided.
  • Automatic project detection and a lightweight registry organize instincts by project and enable promotion between scopes.
  • Instincts can be evolved into skills/commands/agents to automate workflows and improve consistency across tasks.

Quick Start

Install the Continuous Learning v2 tooling and run the observe hooks to start collecting session data.

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 session data to build reusable knowledge?

Hooks-based observation monitors Claude Code sessions, capturing tool usage behaviors and automatically storing them as instinct prompts with full provenance metadata attached.

What are atomic instincts and how do confidence scores work?

Atomic instincts are discrete behavioral patterns extracted from sessions, and confidence scores quantify their reliability to determine readiness for evolving into higher-level skills, commands, or agents.

How do I prevent cross-project contamination when applying learned instincts?

Cross-project contamination is prevented by isolating patterns within project-scoped instincts, ensuring only explicitly promoted universal patterns reach the optional global layer.

Can I evolve Claude Code instincts into more complex automated workflows?

Yes, atomic instincts with high confidence scores can be evolved into skills, commands, or agents to automate workflows and improve consistency across tasks.

What format are instinct artifacts stored in, and how are they organized?

Instinct artifacts are stored as YAML files containing fields like id, trigger, confidence, domain, source, scope, and content, organized within per-project directories using a lightweight projects registry.

Does learning from Claude Code sessions require manual project configuration?

No, automatic project detection organizes instincts by repository and maintains a minimal registry, so you only need to install the tooling and run observe hooks to start collecting data.