Does this skill work with CLAUDE.md?

Yes, it creates a symlink between AGENTS.md and CLAUDE.md so that your instructions are accessible regardless of which naming convention your AI tools prefer.

What are 'file-scoped commands'?

These are commands that run linting or testing on a single file rather than the whole project, which is faster and more reliable for AI agents performing targeted tasks.

What is the primary benefit of using AGENTS.md?

AGENTS.md provides a canonical set of instructions for AI agents, ensuring they use the correct package managers, commands, and conventions specific to your project.

How does it handle different package managers?

The skill analyzes your project for lock files like pnpm-lock.yaml or poetry.lock to automatically determine the correct commands for the agent to use.

Why does this skill enforce a 100-line limit?

Research indicates that AI instruction-following quality degrades as document length increases. Keeping documentation under 100 lines ensures the agent remains focused on high-priority rules.

Agent Instructions Manager

Name: Agent Instructions Manager
Author: lingxling

bylingxling

•

学習とドキュメンテーション

Maintains concise, high-signal documentation for AI agents to improve instruction following and project-specific performance.

The Agent Instructions Manager skill helps developers create and maintain optimized AGENTS.md and CLAUDE.md files. Based on research-backed best practices, it enforces a minimal, high-signal documentation style that limits files to under 100 lines to prevent AI performance degradation. By automatically detecting your project's specific package manager, linter configurations, and file-scoped commands, this skill ensures that AI agents have exactly the context they need without the fluff that leads to errors or hallucinations.

主な機能

01Implements file-scoped commands for efficient, cost-effective testing and linting.

02Automatically detects package managers and toolchains (pnpm, uv, poetry, etc.).

03Standardizes AI commit attribution with mandatory Co-Authored-By headers.

04Eliminates redundant documentation by referencing existing config files instead of duplicating them.

0554 GitHub stars

06Generates minimal AGENTS.md/CLAUDE.md files focused on high-signal instructions.

ユースケース

01Setting up standardized AI collaboration guidelines for a new repository.

02Configuring file-scoped developer commands to speed up AI-driven refactoring tasks.

03Auditing and refactoring bloated agent documentation to improve Claude's accuracy.

主な機能

01Implements file-scoped commands for efficient, cost-effective testing and linting.

02Automatically detects package managers and toolchains (pnpm, uv, poetry, etc.).

03Standardizes AI commit attribution with mandatory Co-Authored-By headers.

04Eliminates redundant documentation by referencing existing config files instead of duplicating them.

0554 GitHub stars

06Generates minimal AGENTS.md/CLAUDE.md files focused on high-signal instructions.

ユースケース

01Setting up standardized AI collaboration guidelines for a new repository.

02Configuring file-scoped developer commands to speed up AI-driven refactoring tasks.

03Auditing and refactoring bloated agent documentation to improve Claude's accuracy.