Is the learning process automatic or manual?

It is designed to be automatic via the Stop hook, but you can also use the `/learn` command to manually trigger pattern extraction during an active session.

Does this skill require a minimum number of messages?

By default, the skill evaluates sessions that have at least 10 messages to ensure there is sufficient context to identify meaningful and reusable patterns.

Can I prevent the skill from learning specific things?

Yes, the skill is fully configurable via `config.json`. You can define ignore patterns for simple typos or one-time fixes that you don't want to be saved as permanent skills.

How does the Continuous Learning skill trigger?

It operates as a 'Stop hook' at the end of every Claude Code session, meaning it only runs once you have finished your current task, ensuring no impact on real-time performance.

Where are the learned skills saved?

Extracted skills are stored locally on your machine, typically in the `~/.claude/skills/learned/` directory, giving you full control over your data.

Continuous Learning for Claude Code

Name: Continuous Learning for Claude Code
Author: Nixdorfer

byNixdorfer

•

生産性とワークフロー

Automatically captures and saves reusable coding patterns, error resolutions, and project-specific conventions from your AI sessions.

The Continuous Learning skill enhances your development workflow by automatically analyzing Claude Code sessions upon completion to identify and extract valuable insights. By monitoring message history, it detects successful error resolutions, user-provided corrections, and complex workarounds, saving them as permanent skills in your local directory. This ensures that Claude evolves alongside your codebase, remembering specific debugging techniques and architectural preferences without requiring manual documentation or repetitive prompting.

主な機能

01Automated session evaluation using non-blocking Stop hooks

02Advanced filtering to ignore trivial typos or one-time external API issues

03Intelligent pattern detection for errors, workarounds, and corrections

04Customizable extraction thresholds and session length requirements

05Local storage of learned skills for persistent cross-session memory

061 GitHub stars

ユースケース

01Automating the documentation of obscure library workarounds discovered during debugging

02Building a personalized knowledge base of effective prompts and user-preferred coding styles

03Capturing project-specific architectural conventions for consistent future code generation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nixdorfer/claudecodetool continuous-learning

For use in Claude.ai and ChatGPT

Download Skill