How does the confidence scoring work in Continuous Learning v2?

Instincts start with a low confidence score and increase as patterns are repeatedly observed or validated by the user, ensuring only reliable behaviors eventually become core habits.

What is the benefit of using hooks over version 1 skills?

Hooks are deterministic and fire 100% of the time during tool use, ensuring every interaction is captured, whereas version 1 relied on probabilistic triggers that could miss subtle patterns.

Is my source code shared with external servers during learning?

No, all observations and instincts are stored locally on your machine. The system only processes patterns, and you maintain full control over what is exported or shared.

Can I share my learned instincts with my development team?

Yes, the system includes dedicated /instinct-export and /instinct-import commands, allowing you to share abstracted patterns and project-specific knowledge with others.

Continuous Learning v2

Name: Continuous Learning v2
Author: x-cmd

byx-cmd

•

Productivity & Workflow

Automates the capture and evolution of AI coding patterns into reusable skills and instincts through deterministic session observation hooks.

Continuous Learning v2 transforms ephemeral Claude Code sessions into a persistent knowledge base by monitoring tool usage and user corrections in real-time. By utilizing deterministic hooks rather than probabilistic triggers, it identifies recurring patterns and creates atomic 'instincts'—weighted behaviors that evolve over time into custom commands and specialized agents. This system is ideal for developers who want Claude to adapt seamlessly to their specific coding styles, architectural preferences, and unique workflow habits without the need for constant manual prompting.

Key Features

01Evolution engine that clusters instincts into specialized agents and commands

029 GitHub stars

03Deterministic session observation via PreToolUse and PostToolUse hooks

04Atomic instinct creation with weighted confidence scoring (0.3 to 0.9)

05Background pattern detection using efficient Haiku model processing

06Privacy-focused local storage with selective export and import capabilities

Use Cases

01Enforcing project-specific functional programming styles and architectural patterns automatically

02Converting recurring error resolutions into permanent instincts to prevent bug regression

03Scaling individual coding preferences into shareable skill sets for team-wide consistency

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add x-cmd/skill continuous-learning-v2

For use in Claude.ai and ChatGPT

Download Skill