What is the benefit of project-scoped instincts?

Project scoping prevents patterns from one project (like React hooks) from incorrectly appearing in another (like a Python backend), keeping Claude's behavior contextually accurate.

How does the system capture my coding patterns?

It uses system-level hooks (PreToolUse and PostToolUse) to observe every tool call and user correction in Claude Code, ensuring no learning opportunities are missed.

Is my data kept private?

Yes, all observations, instincts, and evolved skills are stored locally on your machine within the ~/.claude/homunculus/ directory.

How do instincts evolve into skills?

When multiple related instincts reach a high confidence threshold, you can use the /evolve command to cluster them into complex commands or specialized agents.

Continuous Learning v2.1

Name: Continuous Learning v2.1
Author: saar120

bysaar120

0•

Productividad y Flujo de Trabajo

Transforms coding sessions into a library of evidence-based patterns that evolve into personalized AI skills and commands.

Continuous Learning v2.1 is an advanced architecture for Claude Code that captures atomic 'instincts'—small, learned behaviors with confidence scores—directly from your development sessions. By utilizing deterministic system hooks rather than probabilistic prompts, it ensures 100% of tool interactions are analyzed to identify repeated workflows, user corrections, and project-specific preferences. The system features sophisticated project-scoping to prevent cross-contamination between different codebases, allowing you to maintain distinct patterns for React, Python, or Git workflows while promoting universal best practices to a global library.

Características Principales

01Deterministic Observation Hooks: Employs PreToolUse/PostToolUse hooks for 100% reliable session capture.

02Atomic Instinct Generation: Automatically extracts specific behaviors with 0.3 to 0.9 confidence weighting.

03Automatic Promotion: Identifies high-confidence patterns used across multiple projects for global application.

04Evolution Pipeline: Clusters related instincts to generate full-fledged skills, commands, or specialized agents.

050 GitHub stars

06Project-Specific Scoping: Isolates patterns to specific repositories using Git remote hash detection.

Casos de Uso

01Scaling repetitive multi-step workflows into automated custom commands based on observed developer behavior.

02Building a private, local knowledge base of your personal coding 'best practices' across all repositories.

03Teaching Claude project-specific architectural patterns and naming conventions without manual prompting.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add saar120/skillgate-registry continuous-learning-v2

For use in Claude.ai and ChatGPT

Características Principales

01Deterministic Observation Hooks: Employs PreToolUse/PostToolUse hooks for 100% reliable session capture.

02Atomic Instinct Generation: Automatically extracts specific behaviors with 0.3 to 0.9 confidence weighting.

03Automatic Promotion: Identifies high-confidence patterns used across multiple projects for global application.

04Evolution Pipeline: Clusters related instincts to generate full-fledged skills, commands, or specialized agents.

050 GitHub stars

06Project-Specific Scoping: Isolates patterns to specific repositories using Git remote hash detection.

Casos de Uso

01Scaling repetitive multi-step workflows into automated custom commands based on observed developer behavior.

02Building a private, local knowledge base of your personal coding 'best practices' across all repositories.

03Teaching Claude project-specific architectural patterns and naming conventions without manual prompting.