Does the background observer slow down my terminal?

No, the observation hooks are lightweight shell scripts that run instantly. The heavy lifting of pattern analysis is performed by a background agent (Haiku) that runs at configurable intervals to ensure zero impact on your typing speed.

Will it leak my private code between different projects?

No, version 2.1 introduces strict project-scoping. Patterns learned in one repository are isolated by a unique Git-based hash and are only shared globally if they meet specific 'promotion' criteria or if you manually move them.

What happens when I correct Claude's code?

The system views user corrections as high-signal events. It captures the delta between what Claude did and what you requested, creating a new instinct with evidence so that the same mistake isn't repeated in future sessions.

How does this differ from standard Claude memory?

While standard memory is often ephemeral or context-limited, this system creates structured, evidence-based YAML 'instincts' that are weighted by confidence and can be evolved into permanent commands or shared across projects.

Continuous Learning V2

Name: Continuous Learning V2
Author: affaan-m

byaffaan-m

•

156,032

•

Productivity & Workflow

Automatically transforms Claude Code session observations into reusable instincts and evolved skills through a project-aware background learning system.

Continuous Learning V2 is a sophisticated instinct-based architecture that captures your coding patterns and preferences during Claude Code sessions to build a personalized knowledge base. By utilizing deterministic hooks rather than probabilistic suggestions, it achieves 100% reliability in recording tool usage and user corrections. The system intelligently distinguishes between project-specific rules—such as React patterns or local directory structures—and global best practices, allowing for an ever-improving AI assistant that adapts to your specific workflows across different repositories without cross-project pollution. As instincts gain confidence through repetition, they can be evolved into full-featured commands, skills, or specialized agents.

Key Features

01Project-Scope Isolation: Uses Git repository hashes to keep specialized project patterns from leaking into unrelated environments.

02Automated Skill Evolution: Clusters related instincts to generate production-ready commands and specialized AI agents.

03Global Promotion System: Identifies high-confidence patterns appearing across multiple projects and promotes them to a global library.

04Deterministic Hook Observation: Implements Pre/PostToolUse hooks for 100% reliable session capture and pattern detection.

05Atomic Instinct Generation: Captures specific triggers and actions with weighted confidence scores from 0.3 to 0.9.

06156,032 GitHub stars

Use Cases

01Automating the enforcement of project-specific coding styles and architectural conventions without manual prompting.

02Capturing and preserving personal debugging workflows and error-handling strategies for reuse in future sessions.

03Building a private, evolving library of cross-project AI behaviors that improve Claude's performance based on your historical feedback.

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

Key Features

01Project-Scope Isolation: Uses Git repository hashes to keep specialized project patterns from leaking into unrelated environments.

02Automated Skill Evolution: Clusters related instincts to generate production-ready commands and specialized AI agents.

03Global Promotion System: Identifies high-confidence patterns appearing across multiple projects and promotes them to a global library.

04Deterministic Hook Observation: Implements Pre/PostToolUse hooks for 100% reliable session capture and pattern detection.

05Atomic Instinct Generation: Captures specific triggers and actions with weighted confidence scores from 0.3 to 0.9.

06156,032 GitHub stars

Use Cases

01Automating the enforcement of project-specific coding styles and architectural conventions without manual prompting.

02Capturing and preserving personal debugging workflows and error-handling strategies for reuse in future sessions.

03Building a private, evolving library of cross-project AI behaviors that improve Claude's performance based on your historical feedback.