agents-md

Guides creation, auditing, and revision of AGENTS.md instruction files.

2|Updated Sep 14, 2026
One-click install
npx skills add https://github.com/kwanpham2195/agent-stuff --skill agents-md-kwanpham2195
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agents-md
Source: https://github.com/kwanpham2195/agent-stuff/tree/main/skills/agents-md
Command: npx skills add https://github.com/kwanpham2195/agent-stuff --skill agents-md-kwanpham2195

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AGENTS.md files often become bloated with redundant documentation, task history, and rules that belong in code or configuration. This Skill provides a structured methodology for deciding what belongs in an AGENTS.md file, where each rule should live, and how to keep agent-facing instructions concise and behavior-changing. ## Core Features & Use Cases - Scope Classification: Distinguishes brownfield roots, nested subtrees, greenfield guidance, and personal/global policy files, each with different register and length expectations. - Rule Placement Logic: Routes each rule to its nearest durable home, preferring executable checks and configuration over prose, and reserving AGENTS.md for non-obvious intent, constraints, and hazards. - Length and Content Guardrails: Targets roughly 20 lines for brownfield roots and 1-10 lines for nested files, flagging anything over 40 lines as suspicious. - Use Case: When auditing a repository's AGENTS.md that has grown to 150 lines of directory tours and task history, use this Skill to strip it down to durable, behavior-changing rules and move enforceable checks into configuration. ## Quick Start Review the AGENTS.md file in this repository and revise it according to the agents-md guidance.

Frequently Asked Questions about agents-md

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a good AGENTS.md file for my repository?▼

Start with the first useful rule or scope statement, address the agent as you, and use compact machine-first bullets. Include only non-obvious intent, constraints, hazards, and ownership boundaries, aiming for roughly 20 lines in an ordinary brownfield root.

What should not go in an AGENTS.md file?▼

Exclude stack inventories, directory tours, routine commands, task history, and anything cheaply visible in code or configuration. Prefer executable checks or configuration for enforceable rules, and keep human-facing explanation in README files.

When should I create a nested AGENTS.md file?▼

Create a nested file only for a genuine local delta such as subtree ownership, public API boundaries, generated or frozen areas, required wrappers, or recurring mistakes an agent cannot infer. Keep nested files between 1 and 10 lines.

How long should an AGENTS.md file be?▼

Ordinary brownfield roots should target roughly 20 lines, with more than 40 lines treated as suspicious and over 100 as misplaced documentation. Greenfield and personal files may run longer when they carry durable intent unavailable elsewhere.

Can AGENTS.md files use expressive or personal language?▼

Personal, global, and greenfield files may use first-person prose and distinctive voice when the language carries identity, taste, or decision-bearing intent. Remove generic inspiration and decorative flourishes that change no behavior.