agentleFS
Sign inStart free
andreymudriandreymudri/claude-teammates/AGENTS.md

claude-teammates

AGENTS.md3 starsChanged 10 days agoMITView on GitHub

Instructions for coding agents (Codex, Cursor, and others) working in this repository. Humans: see README.md and CONTRIBUTING.md. Two separate things share this repo: - fleetmates (repo root): a Claude Code plugin plus a CLI (scripts/cli.mjs) that runs a planned change as a fleet of worktree-isolated agents with phase gates. Zero dependencies, runtime and dev. - fleetmates deck (hub/, design docs in docs/deck/): a local web deck over Claude Code sessions. A private npm package with its own dependencies (node-pty, xterm,…

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 1

    1. andreymudri/fleetmates★ 3100% same

    Edited · 0

    None found.

    Wrote their own version · 0

    None found.

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • 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 repository15 other files it gives its agents.

The file

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

4 rules. About 0 are common, 2 are distinctive, and 2 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.

  • 2

    Very specific

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

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

    Use a short TMPDIR for hub tests.

    line 59
  2. Distinctive:

    Before you finish, kill every deckd, deck server, fake claude and headless Chromium you started.

    line 63
  3. Very specific:

    This repo is public: Never write personal names, emails, usernames or real home paths into code, docs, fixtures or test output.

    line 74
  4. Distinctive:

    When Codex orchestrates, dispatch through the CLI rather than Claude Code's Agent tool:

    line 90

≈ “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

Instructions for coding agents (Codex, Cursor, and others) working in this repository. Humans: see README.md and CONTRIBUTING.md.

What is here

Two separate things share this repo:

  • fleetmates (repo root): a Claude Code plugin plus a CLI (scripts/cli.mjs) that runs a planned change as a fleet of worktree-isolated agents with phase gates. Zero dependencies, runtime and dev.
  • fleetmates deck (hub/, design docs in docs/deck/): a local web deck over Claude Code sessions. A private npm package with its own dependencies (node-pty, xterm, ws). The zero-dependency rule does not apply inside hub/.

Current state of the deck

The deck lives on master; the old feat/deck branch is merged and no longer used. Base new deck branches and fleet runs on master. Releases are tagged deck-vX.Y.Z.

Milestones M0 through M5 are implemented, and M6 (Deep research) is a tested release candidate. The exit steps that need the owner (real vault questions, real research runs, the M1 dogfood week) are listed in docs/deck/m1-exit.md through m5-exit.md and docs/deck/m6-progress.md. Start from these files:

  • docs/deck/12-milestones.md: the milestones and what each one covers.
  • docs/deck/README.md: how the design docs are organized.
  • docs/deck/16-platforms.md: Linux, macOS and Windows differences.
  • docs/deck/spikes/m0.md: what M0 built and measured. Section 6 lists early defects with file:line.

Section 6 of the report may be out of date. Before fixing or citing an item from it, check it against the current code: grep for the symbol, and run or mutate the test that is supposed to pin it.

Parts of hub/:

PathContents
hub/deckd/PTY daemon on a Unix socket. The protocol is in docs/deck/05-api.md section 5
hub/bin/fm.mjsfm claude and fm attach
hub/server/screen/Screen parsers
hub/server/adapters/scribed.mjsscribed client
hub/test/fixtures/Captured Claude Code 2.1.282 hook and screen fixtures, already redacted

Commands

Root (fleetmates):

npm test

Hub:

npm ci --prefix hub
mkdir -p /tmp/hx && TMPDIR=/tmp/hx npm --prefix hub test
TMPDIR=/tmp/hx npm --prefix hub run perf    # keystroke echo through Focus; headless /usr/bin/chromium, override with CHROMIUM_PATH

Use a short TMPDIR for hub tests. Unix socket paths are limited to about 108 bytes, and a long temp dir makes the socket tests fail with errors that look like real bugs. CI runs the root suite, and .github/workflows/deck.yml runs the hub suite on Linux, macOS and Windows.

Before you finish, kill every deckd, deck server, fake claude and headless Chromium you started. From the repository root this command must print nothing:

pgrep -af "$PWD/hub/(deckd|server)/main\.mjs|fake-claude\.mjs"

Rules

  • Commits: single line, commitlint style, English, scope deck for hub work (for example fix(deck): ...). The configured git user is the only author. Never add Co-Authored-By, session links, "Generated with" lines, or any other tool attribution, in commits, PR bodies or tags.
  • This repo is public. Never write personal names, emails, usernames or real home paths into code, docs, fixtures or test output. Use placeholders such as /home/you. Fixture redaction lives in hub/test/capture/capture-cc.mjs. Recapturing fixtures starts a real Claude Code session, so the owner must authorize each one.
  • Docs in docs/deck/: English, plain prose, no em dash character, placeholders for personal paths.
  • Tests must be able to fail. For each new test, break the code it covers, watch the test fail, then restore the code. A mutation that leaves the suite green means the test does not pin what its name says. See CONTRIBUTING.md.
  • Never modify the fleetmates gate or plan to make your own work pass. That includes fleetmates.gate.json, the plans in docs/plans/, and anything under .fleetmates/ (gitignored run state). When a fleet run is active, the CLI decides which files each task may touch.

Running a fleet from Codex

When Codex orchestrates, dispatch through the CLI rather than Claude Code's Agent tool:

node scripts/cli.mjs dispatch --run <id> --phase <n> --harness codex --root .
node scripts/cli.mjs dispatch-reviews --run <id> --phase <n> --harness codex --root .
node scripts/cli.mjs dispatch-integrator --run <id> --phase <n> --harness codex --root .

The gate is the same one Claude Code uses:

node scripts/cli.mjs gate --run <id> --plan <path> --base master --root . --phase <n>

A full multi-phase fleet run driven from Codex has not been done yet. Check each step's output before relying on it.

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.

Check your setup, free