agents-md_skill

This skill helps you create, verify, and improve AGENTS.md files with progressive disclosure and concise root guidance.

TypeScript

0

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill dimitrigilbert/ai-skills --skill agents-md

SKILL.md10.2 KB

Overview

This skill is an expert AGENTS.md file assistant for creating, verifying, and refactoring AGENTS.md documents. I help produce minimal, task-focused root files, design progressive-disclosure hierarchies, and verify existing AGENTS.md for bloat, contradictions, and stale information. Use it to make agent prompts reliable and portable across models and repos.

How this skill works

I inspect an AGENTS.md (or your repo) for essentials: a one-sentence project description, package manager, non-standard build/typecheck commands, and links to progressive-disclosure files. I run a checklist for bloat indicators, contradiction detection, and stale-path risks, then propose a refactor: a tiny root AGENTS.md plus focused domain files and a suggested folder layout. I can also generate or patch a minimal AGENTS.md from answers to a short set of clarifying questions.

When to use it

Creating a new AGENTS.md for a single-package project
Refactoring a bloated or contradictory AGENTS.md
Verifying an AGENTS.md before publishing or CI use
Adding nested AGENTS.md for a monorepo package
Converting detailed rules into progressive-disclosure docs

Best practices

Start with a one-sentence project description that acts as a role prompt
Keep root AGENTS.md minimal — include only what every task needs
Prefer links to focused docs (TYPESCRIPT.md, TESTING.md) rather than embedding all rules
Describe capabilities and patterns, not exact file paths
Use conversational language, avoid absolute terms like ALWAYS or MUST

Example use cases

Generate a minimal AGENTS.md from answers about repo type, stack, package manager, and build/test commands
Verify an existing AGENTS.md and return a checklist of issues: bloat, contradictions, stale paths
Refactor a long AGENTS.md into: root AGENTS.md + docs/TYPESCRIPT.md + docs/TESTING.md with suggested contents
Create package-level AGENTS.md for a monorepo package that merges with root guidelines
Suggest CI-friendly AGENTS.md content that keeps tokens low and reduces hallucination risk

FAQ

Root should contain only what every agent task needs: one-sentence description, package manager (if not npm), and non-standard build/typecheck commands. Move conventions, examples, and lengthy rules into linked domain files.

How long should AGENTS.md be?

Aim for under 50 lines at the root. If content exceeds that, split into progressive-disclosure files to avoid token waste and stale context.