agentleFS
ditomaxditomax/build/AGENTS.md

build

AGENTS.md1 starsChanged 29 days agoMITView on GitHub

You are working inside a build workspace. build turns a maquette brief, an idea register or shortlist entry, or a problem statement into a product through concept documents that the code is synthesized from, in four stages per feature with three human gates. The person you are talking to may be a product owner, not a developer. They do not need to know stage names, files or commands — you do. Before anything else. If this file is not in…

  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository6 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read RULES.md (binding for everything you do here) and keep MANIFEST.md at hand — it is the process the stages implement.

    line 9
  2. Very specific:

    Read skills/build/SKILL.md and act as the Director described there.

    line 10
  3. Very specific:

    Exception: if the user only asks about these files — what START.md or this file says, whether the guide is clear — answer the question and …

    line 11
  4. Very specific:

    If profile/profile.md exists, read it — it carries this customer's constraints and names the other profile files (RULES §8).

    line 13
  5. Very specific:

    Write planning files only inside the work root (builds/<project>/ or planning/build/) and code only under the code_root recorded in …

    line 36
  6. Very specific:

    Never write into a maquette or idea folder: 60-brief.md, vcode/, 10-register.md and 10-shortlist.md are read-only.

    line 37
  7. Distinctive:

    Never install packages, add dependencies the concept does not name, run git init, or open network connections without saying so first and …

    line 38
  8. Distinctive:

    No telemetry, no hidden files, no state outside the work root.

    line 39

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

AGENTS.md — build workspace

You are working inside a build workspace. build turns a maquette brief, an idea register or shortlist entry, or a problem statement into a product through concept documents that the code is synthesized from, in four stages per feature with three human gates. The person you are talking to may be a product owner, not a developer. They do not need to know stage names, files or commands — you do.

Before anything else. If this file is not in your working directory but in a subfolder, that subfolder is the workspace — work from there and never write outside it. If you are reading this without a local copy (on GitHub), get one first: README.md § For agents. Inside a project folder (planning/suite/build/) this file is not the entry point — planning/AGENTS.md is.

On every session start

  1. Read RULES.md (binding for everything you do here) and keep MANIFEST.md at hand — it is the process the stages implement.
  2. Read skills/build/SKILL.md and act as the Director described there. Do not wait for a slash command: if the user says "start", "next", "redo", "stop" (or the same in their language), or describes what to build, that is your cue. Exception: if the user only asks about these files — what START.md or this file says, whether the guide is clear — answer the question and do not start the workflow (no Director greeting, no workspace).
  3. Mention the version from VERSION once, in half a sentence, at the first greeting.
  4. If profile/profile.md exists, read it — it carries this customer's constraints and names the other profile files (RULES §8). profile/README.md only explains the folder.

Layout

AGENTS.md / CLAUDE.md   this bootstrap (identical content)
START.md                the three steps for the human
RULES.md                shared rules for all stages, working principles, definition of done
MANIFEST.md             the Concept Document Authoring Manifest v8
README.md               what build is, for humans
checklists/             requirement and architecture-consistency checklists
templates/              one template per result file and register — the binding content definition
skills/                 the Director and the four stage skills
profile/                optional customer constraints (empty = core defaults)
builds/                 the user's work in the standalone form: one subfolder per project

Stage skills live at skills/build-<stage>/SKILL.md; the Director calls them by reading those files. The workspace root is <suite> in the skills' wording.

Inside a project the same suite sits at <project>/planning/suite/build/; then the work lives in <project>/planning/build/, the profile in <project>/planning/profile/, briefs in <project>/planning/maquette/<x>/60-brief.md, register and shortlist in <project>/planning/idea/, and the product code in <project>/ outside planning/ (RULES §2, §9). The Director tells the two layouts apart by itself.

Hard limits in this workspace

  • Write planning files only inside the work root (builds/<project>/ or planning/build/) and code only under the code_root recorded in 00-build.md — never modify templates/, skills/, checklists/, profile/, RULES.md, MANIFEST.md or this file. If you think a template is wrong, tell the user; they report it upstream.
  • Never write into a maquette or idea folder: 60-brief.md, vcode/, 10-register.md and 10-shortlist.md are read-only.
  • Never install packages, add dependencies the concept does not name, run git init, or open network connections without saying so first and getting a yes. If a git repository already exists, commit only as RULES §6 defines — frozen states, never progress.
  • No telemetry, no hidden files, no state outside the work root.
  • Talk in the user's language (German → informal "du"). Write result files in the language set in 00-build.md; keep template headings, requirement text, IDs, code and commit messages in English.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.