Agent Instruction Manager

概要

This skill provides a standardized framework for managing AI agent instruction files like CLAUDE.md and AGENTS.md. It implements advanced patterns such as progressive disclosure and minimal root files to prevent token bloat while ensuring Claude has exactly the context it needs for specific tasks. Whether you are initializing a new project, refactoring a 'ball of mud' instruction file, or syncing multi-tool configurations in a monorepo, this skill ensures your AI interactions remain efficient, accurate, and scalable by describing capabilities rather than fragile file paths.

主な機能

• Automated creation and initialization of CLAUDE.md and AGENTS.md files
• Implementation of progressive disclosure patterns for multi-layered documentation
• Synchronization tools to keep multiple AI agent configuration files in sync
• Monorepo support with nested package-level instruction strategies
• 0 GitHub stars
• Audit and refactor workflows to eliminate instruction bloat and redundant rules

ユースケース

• Setting up a new repository with standardized AI agent guidelines and build commands
• Managing complex monorepos where different packages require unique context without duplicating global rules
• Cleaning up a massive CLAUDE.md file that has become too expensive or confusing for the agent