$ AGENTS.md + mvp-system-design skill → free, MIT license

Ship stable, tested, secure apps
in weeks with Codex.

A free playbook for OpenAI Codex: short rules for your AGENTS.md, plus a skill with the 16 chapters Codex follows. Your app goes from an empty repo to production and keeps running after launch.

Get it free → What goes in AGENTS.md
16chapters, repo to production
99%test coverage, enforced
$0free forever, MIT license
// short answer

What to put in AGENTS.md for Codex

Keep AGENTS.md short, and about how Codex should work: think before coding, change only what was asked, keep it simple, prove every change with a test. Codex reads it at the start of every run, and by default stops after 32 KiB. So the long rules (stack, tests, security, release) go in a skill that Codex opens when a task needs it.

3 levels

Home, repo, folder

Codex reads ~/.codex/AGENTS.md for every project, then the one at your repo root, then one per folder down to where you run it. The closest file comes last, so it has the last word.

32 KiB

A size limit

By default Codex stops adding files after 32 KiB. The playbook's 16 chapters come to about 85 KB, more than two and a half times that, so they ship as a skill instead.

once per run

Read when a run starts

Codex builds its instructions when a run starts. Changed AGENTS.md? Start Codex again. An AGENTS.override.md in the same folder replaces it.

Source: OpenAI's Codex docs on AGENTS.md, checked 7 Nov 2026. The full guide, with a template: what to put in AGENTS.md.

// the AGENTS.md rules

What the rules tell Codex

They come with the playbook, ready to copy into your AGENTS.md. About 6 KB, roughly a sixth of what Codex reads.

  • Follow the playbook. Read its chapter before working in an area, and ask before using tech outside its stack.
  • Think before coding. State assumptions, name what's unclear, and ask instead of guessing.
  • Keep it simple. The least code that solves the problem, and nothing nobody asked for.
  • Change only what's asked. Touch what the task needs, clean up only its own mess.
  • Prove it works. Turn every task into a check, such as a test, and loop until it passes.
  • You decide. Plain words about money, users and risk, and you make the call.
  • Git, always. Every change on a branch and through a pull request, never straight to main.
// how to use it with codex

Set it up in Codex in four steps

In the Codex CLI, the IDE extension or the app; for a new project or one you've already started.

  1. step 1

    Get the playbook

    Leave your email below. The link opens right away.

  2. step 2

    Copy the rules into AGENTS.md

    At your repo root, or in ~/.codex/AGENTS.md to use them in every project.

  3. step 3

    Hand it to Codex

    Paste the one line from the playbook into Codex. It adds the playbook as a skill; type $mvp-system-design to call it, or let Codex pick it when a task fits.

  4. step 4

    Ship to production

    Codex builds in the playbook's order, from the first commit to launch.

// codex or claude code?

Codex or Claude Code? Same playbook

The playbook doesn't depend on the agent. Codex reads AGENTS.md and keeps skills in .agents/skills; Claude Code reads CLAUDE.md and keeps them in .claude/skills. The rules and the 16 chapters are the same, so a team can use both. Building with Claude Code? Here's the playbook as a Claude Code skill.

// what's inside the skill

16 chapters, from an empty repo to production

Plain rules: what to build, with what, and in which order.

  1. 01system design
  2. 02stack
  3. 03folders
  4. 04logs
  5. 05dev
  6. 06stage
  7. 07prod
  8. 08release
  9. 09cloud
  10. 10alerts
  11. 11tasks
  12. 12tests
  13. 13code quality
  14. 14UI
  15. 15security
  16. 16rate limits
  17. +AGENTS.md rules for Codex
// free resource

Get the playbook

Free, no catch. You'll get the link as soon as you send the form.

  • All 16 chapters, free forever
  • Rules for your AGENTS.md, ready for Codex
  • Use it on any project, commercial included
// get access

You'll also get occasional emails from me. Unsubscribe anytime. Privacy

// faq

Questions

Does Codex read AGENTS.md by itself?

Yes. At the start of every run, Codex reads ~/.codex/AGENTS.md, then the AGENTS.md files from your repo root down to the folder you're in. There's nothing to switch on.

Why not paste the whole playbook into AGENTS.md?

By default Codex stops reading AGENTS.md files after 32 KiB, and the 16 chapters are about 85 KB. The short rules go in AGENTS.md; the chapters stay in the skill, which Codex opens when a task needs it.

Where does Codex keep the skill?

In an .agents/skills folder: inside your repo for one project, or in your home folder for all of them. Type $ in Codex to call a skill by name.

Which model should I use?

The playbook suggests GPT-6.1 Sol at medium effort or higher.

Is it really free?

Yes. Fill in the form on this page and it's yours, under the MIT license. Use it on any project, including commercial ones.

Do I need to be a developer?

No, but you should be comfortable in a terminal and with Codex. Codex does the heavy lifting; the playbook tells it what good looks like.

I've already started my app. Is it too late?

No. Ask Codex to align the project with the playbook. It closes the gaps step by step instead of rewriting everything.

Can you help me apply it?

Yes. In ProCode Extended you apply it to your own app with me and finish with it live in production. Or book a consultation.