Borrowing it
Nothing to install: this file belongs to miikkij/aimeat-protocol. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/miikkij/aimeat-protocol/main/AGENTS.mdgit clone --depth 1 https://github.com/miikkij/aimeat-protocolWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/miikkij/aimeat-protocol/agents-md)<a href="https://agentmods.dev/instructions/miikkij/aimeat-protocol/agents-md"><img src="https://agentmods.dev/badge/instructions/miikkij/aimeat-protocol/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/miikkij/aimeat-protocol/agents-md"><img src="https://agentmods.dev/badge/instructions/miikkij/aimeat-protocol/agents-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.02103 | $0.02103 |
| Opus 5.5 | $0.00841 | $0.00841 |
| Sonnet 5.5 | $0.00421 | $0.00421 |
| Haiku 4.5 | $0.00210 | $0.00210 |
Grade A, and why
aimeat-protocol AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 18d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 105 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
For coding agents that do not read CLAUDE.md: Codex, Cursor, Copilot, Gemini CLI and the rest.
Claude Code reads CLAUDE.md. That file is the source, and this one exists so that you arrive with
the same rules. It was written on 2026-09-19, after a model working in this repository without any
of them built the wrong thing three times in a row.
Read these first, in this order
CLAUDE.md, in full. What the project is for, how Jouni works, the gates, where knowledge lives. Everything below is a pointer into it or a rule that bites before you would find it..claude/rules/*.md. Rules for one kind of code. Thepathslist at the top of each file says which files it covers. Claude Code loads them by itself when it opens a matching file; you have to open the ones that match what you are about to touch. The table inCLAUDE.md("Rules that load by path") says which is which..claude/skills/<name>/SKILL.md. How one kind of task is done here. Read the one that matches the task BEFORE you start:aimeat-app-building: building, publishing or changing an app on AIMEAT.aimeat-frontend-verify: any change to a page, and how it is checked in a real browser.aimeat-design-language: fonts, colours, shapes, and what a change to one of them reaches.aimeat-writing: any text a person reads, in English, Finnish or Spanish.aimeat-library-authoring: shared code (served libs, cortex libs, extensions, packs).aimeat-imagery: any picture (scripts/gen_image.py, run withuv run python).aimeat-organism-records,aimeat-pages-links,aimeat-video: as their names say.
- The node, over MCP.
aimeat_handbook_getwith no arguments is the node's operating guide and lists its skills by situation;aimeat_skill_getloads one. What the node says wins over what you remember.
Building an app: the rules that were broken
- A new app is built on the Atelier track, forked from a genre. Load the skill
node:aimeat-app-builder-atelierand read its specification:aimeat_handbook_get { tier: "build-app-atelier" }, in parts:genre,libraries,bookandpatternsbefore any code. The Classic track is for an app that is already Classic, or when Jouni asks for it by name. "Quicker to start" is not a reason, and a track is never changed in the middle of a build. - Ask the level first. A quick prototype (the kit as it comes, no styling), an ordinary page
(a ready layout and look from the Design Book), or the finest (a forked genre, the Book's parts,
components of its own, each written down with its reason). All three are Atelier and
bilingual, so a prototype is raised later without starting over. The page states it in
<meta name="aimeat-level" content="proto|plain|fine">. Never pick a lower level silently. - Propose before you build, and wait. On the finest level: two or three genres by name with why each fits, the languages, who signs in, the first screen in a sentence. Jouni chooses. He must not find out what you chose by looking at the result.
- English and Finnish, always, unless he asks for one. The language of the conversation is not the app's language list. Every string goes through a dictionary; the node's own sign-in bar carries the switch. Never write a header control or a language switch of your own.
- The working screen is a mosaic inside the genre, and its arrangement comes from the Design
Book. The genre is the frame; the list, the numbers, the form and the history are
AIMEAT.atelier.mosaic({ target, sources, fallback })mounted into one element of the genre page, withfallbacktaken from a Book fill. A genre page with no mosaic cannot use one part of the Book. The page names the parts (<meta name="aimeat-book-parts">) and carries the layout as data (<script type="application/json" id="aimeat-layout">); the publish counts the parts and stores the layout, so nothing is called afterwards. The proposal owes one line: "From the Design Book I take: …" or "… I take nothing, because …". - Write down why, as you choose, and put nothing into the Design Book because a build is
finished. The page carries
<script type="application/json" id="aimeat-build-notes">withtook,passedandmade, one true sentence of why each; the publish keeps them with the version. Jouni rebuilt one app three times and liked the third: only an app its owner says turned out well teaches the Book anything. When he says so, record it withaimeat_designbook_keep { filename }, and then offer what the app made by hand as Design Bookcomponentparts (markup and a stylesheet under one class prefix, colours asvar(--ak-…)tokens, no script), each with your honest judgement:generalorspecial, and why. A general one is published by itself. Build hand-made pieces in that shape from the start. - The genre is the look. A look preset with components stacked in it, under a genre's name, is the default page every app looks like, and it is what gets sent back.
- Judge what it looks like, not only that it works. Passing tests, contrast and page width do not make a page good. Put the result beside the genre it was forked from, at 390 and 1440 px, in both themes and both languages, and ask whether it holds up. A broken header in your own screenshot is a reason to stop, whatever else passed.
- Look for a browser before you say you have none, and never let its absence hold a publish
back. List your tools and search them for
playwright,browserandchrome: this machine has a Playwright MCP server driving Chrome, and a session that has it and does not look is the failure seen three builds in a row. With no browser, publish anyway, hand over the address, and say in one line what you could not check yourself. - Done means all of what was agreed. Check every requirement against the result before you say done. Jouni is not the quality check you left out.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 18d ago Changed · +15 lines · +355 tokens per session cb6609f8d89a
- 1mo ago Changed · +9 lines · +209 tokens per session da47697a7b30
- 1mo ago First seen · 81 lines · 1,539 tokens per session scan A 800c64be516b
aimeat-protocol AGENTS.md is an instructions file published in the GitHub repository miikkij/aimeat-protocol (10 stars, last pushed today), licensed MIT. It adds 2,103 tokens to every session, about $0.0084 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-20.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deno copilot-instructions.md
Copilot instructions for denoland/deno, covering deno development guide for github copilot, network access, high level overview, key directories and key files to understand first.