continuous-learning-v2

Capture Claude Code session tool events and convert them into confidence-weighted YAML instinct files.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/JGcode3675/gcs-os --skill continuous-learning-v2-jgcode3675
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: continuous-learning-v2
Source: https://github.com/JGcode3675/gcs-os/tree/main/home/james/claude-config/skills/continuous-learning-v2
Command: npx skills add https://github.com/JGcode3675/gcs-os --skill continuous-learning-v2-jgcode3675

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

Automatically capture deterministic tool events from Claude Code sessions and convert recurring behaviors into small, confidence-scored "instincts" so user preferences and repeated workflows become actionable automation without manual rule authoring.

Core Features & Use Cases

  • Hooks-based observation that records PreToolUse and PostToolUse events to a local observations.jsonl store.
  • Background observer agent (haiku model by default) that detects patterns, assigns confidence scores, and writes atomic YAML instinct files with evidence and domain tags.
  • CLI utilities to inspect instincts, import/export shared instincts, and cluster/evolve instincts into skills, commands, or specialist agents.
  • Use cases: automatically prefer a coding style observed during sessions, surface a repeatable debugging workflow as a command, or share team conventions via exported instincts.

Quick Start

Enable the PreToolUse and PostToolUse observation hooks, start the observer agent, and run the instinct status command to begin collecting and reviewing 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 automatically capture Claude Code session events and convert them into reusable rules?

You can capture session events by enabling PreToolUse and PostToolUse hooks to record tool usage to a local observations.jsonl store, which a background observer agent converts into confidence-scored YAML instinct files.

What are confidence-weighted instincts and how do they encode developer workflows?

Confidence-weighted instincts are atomic YAML files that encode user patterns by recording a trigger, confidence score, domain, and evidence, turning repeated developer workflows and style preferences into actionable automation.

How do I share team coding conventions using exported YAML instinct files?

You can share team conventions by using the CLI import and export utilities to transfer YAML instinct files, allowing other developers to apply observed coding styles and debugging workflows to their own sessions.

Can I turn repeated debugging workflows observed in Claude Code into custom commands?

Yes, the CLI utilities support clustering observed instincts to evolve them into new skills, commands, or specialist agents, allowing repeated debugging sequences to become reusable automation.

Does the observer agent require specific hooks to detect session patterns?

Yes, the observer agent relies on Claude Code sessions emitting PreToolUse and PostToolUse hooks to capture deterministic tool events, which are then analyzed to detect patterns and assign confidence scores.

What is the best way to automatically prefer a coding style observed during sessions without manual rule authoring?

The best way is to enable session observation hooks and let the background observer agent detect your recurring style preferences, automatically generating atomic confidence-scored instincts without manual rule writing.