Skip to content
Back to skills

Agents Md Guidelines

ASecurity

Guidelines for writing small, stable AGENTS.md files. Use when creating, refactoring, or reviewing AGENTS.md.

  • 57 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 17, 2026
code-qualitytestingdebuggingrefactoring

Security analysis

A100/100

Scanned October 17, 2026

npx -y skills add diegosouzapw/awesome-omni-skill --skill agents-md-guidelines --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Agents Md Guidelines?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Agents Md Guidelines
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/diegosouzapw-agents-md-guidelines/badge)](https://www.skillsdirectory.com/skills/diegosouzapw-agents-md-guidelines)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: agents-md-guidelines
description: Guidelines for writing small, stable AGENTS.md files. Use when creating, refactoring, or reviewing AGENTS.md.
---

# AGENTS.md Guidelines

## Core Principles

- Keep root AGENTS.md as small as possible; it loads on every request.
- Use progressive disclosure: link to deeper docs for details.
- Prefer stable concepts over volatile paths or time-sensitive notes.
- Avoid “ball of mud” growth; don’t add rules for every mishap.
- Keep instructions concise; assume the agent already knows basics.

## Essentials Only

Root AGENTS.md should usually include:
- One-sentence project description
- Package manager if non-default
- Non-standard build/typecheck commands
- Pointers to deeper docs

## Progressive Disclosure

- Move language, testing, and workflow rules into separate docs.
- Link to those docs from AGENTS.md.
- Keep guidance domain-based, not file-path based.
- Avoid deep reference chains; link one level deep when possible.

## Staleness Rules

- Avoid exact file paths unless they are stable interfaces.
- Avoid dates or time-based instructions.
- Prefer capability descriptions over structure hints.

## Monorepo Guidance

- Use root AGENTS.md for shared rules.
- Use package-level AGENTS.md for package-specific conventions.
- Keep each level focused and minimal.

## Instruction Design

- Start with a TL;DR checklist for fast scanning.
- Include clear triggers and skip cases to prevent overuse.
- Provide ordered steps with explicit outputs.
- Use strict format rules with correct/incorrect examples.
- Add a decision tree for edge cases when needed.
- Include validation and debugging tips to reduce failure loops.
- Make naming conventions explicit.
- Define archive or completion criteria.
- Keep terminology consistent.

## Quality Checklist

- Description includes what + when, third-person.
- Root AGENTS.md is short and link-heavy.
- TL;DR + triggers + steps are present when needed.
- No duplicated guidance across files.
- Terminology is consistent.
- Examples are concrete and brief.
- Instructions fit an "instruction budget".

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.