How do I trigger the self-updating process?

You can trigger it naturally by saying phrases like 'remember this for next time', 'update the rules based on that error', or 'record this feedback'.

What files does this skill typically modify?

It primarily updates CLAUDE.md for user preferences and methodology files located in the tips/ directory of your repository.

Does it just add text to the bottom of my files?

No. The skill is designed to read the entire target file, understand the structure, and integrate new rules into the most appropriate section to keep documentation clean.

Do I need to explain what went wrong for it to work?

The skill is optimized to 'backtrack' through the conversation history to automatically identify the failure or the correction, though you can provide specific details if you prefer.

Self-Updating Agent Manager

Name: Self-Updating Agent Manager
Author: majiayu000

bymajiayu000

•

Productivity & Workflow

Enables Claude to learn from errors and feedback by automatically updating internal rules and methodology documentation.

This meta-skill empowers Claude to evolve through continuous feedback loops and automated documentation maintenance. When a user corrects an error or provides a new instruction, the skill triggers an autonomous workflow that analyzes conversation context, identifies the root cause, and intelligently integrates new rules into existing documentation like CLAUDE.md or specific methodology files. Unlike simple logging, it refines the core logic of the agent's behavior by placing rules in their proper structural context, ensuring that lessons learned are applied to all future tasks without repeating the same mistakes.

Key Features

0164 GitHub stars

02Methodology Evolution: Specifically designed to refine 'tips' and workflow guides based on real-world execution results.

03Intelligent Rule Integration: Merges updates into the relevant sections of documentation rather than just appending to the end.

04Contextual Error Analysis: Automatically retraces interaction history to pinpoint failure points without user explanation.

05Multi-File Synchronization: Updates repository manifests and routing tables to maintain system-wide consistency.

06Event Logging: Maintains a separate factual log of incidents to track why and when rules were updated.

Use Cases

01Refining complex workflows like video processing or code generation after a logic failure.

02Personalizing agent behavior by recording specific user preferences and stylistic choices.

03Codifying 'lessons learned' during long-term projects to prevent regressions in AI performance.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add majiayu000/claude-skill-registry 自更新

For use in Claude.ai and ChatGPT

Download Skill