AGENTS.md
SkillFiles & storageAuthor and maintain AGENTS.md files — the open, tool-agnostic format for giving coding agents project context. Use when: creating AGENTS.md; writing an agent instructions file; root or nested AGENTS.md for monorepos; migrating CLAUDE.md, AGENT.md, .cursor/rules, or copilot-instructions.md to AGENTS.md; greenfield project setup before code exists; brownfield onboarding for an existing codebase; documenting build/test/lint commands, code style, PR/commit conventions, and security considerations for agents; AGENTS.md discovery and precedence rules. Works for any language, framework, or project size. DO NOT USE FOR: authoring SKILL.md skill files (use create-skill); APM package/plugin authoring (use apm-producer); language-specific style rules (use language skills e.g. csharp-best-practices).
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the AGENTS.md skill
What this skill tells your AI
The instructions your AI receives, as published by netfabric/netfabric.numerics in .agents/skills/agents-md/SKILL.md and read by ahel’s review.
A README for agents: plain Markdown, no required fields, read by 20+ coding agents (Codex, Cursor, Copilot, Aider, Gemini CLI, Windsurf, Devin, and more).
Quick Facts
| Fact | Detail |
|---|---|
| Location | AGENTS.md at repo root; additional copies nested per package/subproject |
| Discovery | Agent reads the nearest AGENTS.md up the directory tree from the file it's editing |
| Precedence | Closest file to the edited file wins; explicit chat/user instructions override everything |
| Format | Standard Markdown, any headings, no schema to validate |
| Execution | Agents run commands listed in it (tests, lint) and try to fix failures before finishing |
Workflow Picker
flowchart TD
A[Writing AGENTS.md] --> B{Code, CI, or docs already exist?}
B -- No --> C[Greenfield workflow]
B -- Yes --> D[Brownfield workflow]
C --> E{≥2 packages?}
D --> E
E -- Yes --> F[Nest AGENTS.md per package]
E -- No --> G[Single root AGENTS.md]
| Situation | Use |
|---|---|
| No code yet / scaffolding a new repo | Greenfield workflow |
| Existing codebase, README/CI/CLAUDE.md already exist | Brownfield workflow |
| Monorepo with multiple packages | Nesting rules |
Migrating from AGENT.md, CLAUDE.md, .cursor/rules, copilot-instructions.md | Legacy migration |
Core Principle
Never hardcode language, framework, or tool assumptions into an AGENTS.md — every instruction must reflect the target repo's actual, verified stack and commands, not a generic template left unfilled.
AGENTS.md is still a Markdown file: run markdown-best-practices over the draft before treating it as finished (heading spacing, table/list consistency, fenced-code language tags).
Reference Files
| File | Load When |
|---|---|
| references/anatomy-and-discovery.md | Deciding where files live, how precedence works, or migrating from a legacy agent file |
| references/authoring-workflow.md | Writing a new AGENTS.md for a greenfield or brownfield project |
| references/writing-style.md | Wording and formatting an AGENTS.md section for density and scannability |
| references/quality-and-security.md | Reviewing content quality, verifying commands, or covering security considerations |
Signals
- GitHub stars
- 36
- Forks
- 1
- Last commit
- Aug 2026
Advanced
- Catalog kind
- skill
- Gateway key
agents-md-netfabric- Source
- github.com/netfabric/netfabric.numerics