How does v2 differ from the original version?

Version 2 moves from probabilistic skill-based observation to 100% deterministic hooks, introduces background processing with the Haiku model, and adds confidence-weighted scoring for learned behaviors.

How do instincts evolve into skills?

When multiple related instincts reach a high confidence threshold, the '/evolve' command clusters them together to generate specialized agents or permanent skill files.

What is an 'instinct' in Continuous Learning v2?

An instinct is an atomic, learned behavior—such as a preference for functional programming—that is assigned a confidence score based on how frequently Claude observes you using or confirming that pattern.

Does this skill share my private code?

No. All session observations are stored locally on your machine. Only the distilled 'instincts' can be exported, and you have full control over what is shared.

Continuous Learning v2

Name: Continuous Learning v2
Author: affaan-m

byaffaan-m

•

43,117

•

Productivity & Workflow

Automates the capture and evolution of coding patterns into persistent instincts, skills, and specialized agents.

Continuous Learning v2 is a sophisticated behavioral adaptation framework for Claude Code that transforms session interactions into reusable knowledge. By leveraging deterministic PreToolUse and PostToolUse hooks, it ensures 100% reliable observation of your prompts and tool calls. These observations are analyzed by a background agent to identify user corrections, stylistic preferences, and repeated workflows, distilling them into confidence-weighted 'instincts'. Over time, these atomic behaviors evolve into complex skills and specialized agents, creating a truly personalized AI developer experience that grows smarter with every interaction.

Key Features

01Background pattern detection for error resolutions and workflow preferences

0243,117 GitHub stars

03Automated evolution of clusters of instincts into full skills, commands, and agents

04Local-first privacy architecture with granular export/import controls for instincts

05Deterministic session observation via 100% reliable PreToolUse and PostToolUse hooks

06Atomic instinct modeling with dynamic confidence scoring from 0.3 to 0.9

Use Cases

01Personalizing Claude to follow specific architectural patterns or library preferences without repetitive prompting

02Capturing and memorializing specific bug fixes to prevent the AI from regressing on similar issues in the future

03Streamlining complex, multi-step developer workflows into one-line custom commands based on observed behavior

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add affaan-m/everything-claude-code continuous-learning-v2

For use in Claude.ai and ChatGPT

Download Skill