agentleFS
Git-on-my-levelGit-on-my-level/codex-autorunner/AGENTS.md

codex-autorunner

AGENTS.md875 starsChanged 8 months agoMITView on GitHub

This repo dogfoods codex-autorunner. Start here, then prefer the nearest nested AGENTS.md over repeating repo-wide rules.

  • 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 31,439 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 repository3 other files it gives its agents.

The file

Show its rules and text

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

13 rules. About 0 are common, 3 are distinctive, and 10 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 10

    Very specific

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

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

    Read .commandcode/taste/taste.md for project-specific agent taste and preferences.

    line 8
  2. Very specific:

    If the task mentions tickets, contextspace, dispatch, or .codex-autorunner/, also read .codex-autorunner/ABOUT_CAR.md.

    line 9
  3. Very specific:

    If you need runtime setup details, read .codex-autorunner/DESTINATION_QUICKSTART.md.

    line 10
  4. Distinctive:

    Keep the layered split intact: protocol-agnostic engine, adapters, surfaces.

    line 16
  5. Distinctive:

    Prefer deterministic behavior, explicit config, small diffs, and observable failures.

    line 17
  6. Very specific:

    Update tickets only under .codex-autorunner/tickets/.

    line 31
  7. Very specific:

    Update context docs only under .codex-autorunner/contextspace/.

    line 32
  8. Distinctive:

    Do not create duplicate ticket or context files elsewhere in the repo.

    line 33
  9. Very specific:

    Treat .codex-autorunner/ as intentional structure even when generated or gitignored.

    line 34
  10. Very specific:

    Prefer make targets for common checks.

    line 40
  11. Very specific:

    Use gh for GitHub operations when possible.

    line 41
  12. Very specific:

    Use a 300 second timeout for git commit.

    line 42
  13. Very specific:

    Do not restart launchd services or run refresh scripts yourself.

    line 43

≈ “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

Codex Autorunner - Agent Guide

This repo dogfoods codex-autorunner. Start here, then prefer the nearest nested AGENTS.md over repeating repo-wide rules.

Start Here

  • Read docs/car_constitution/61_AGENT_CHEATSHEET.md first.
  • Read .commandcode/taste/taste.md for project-specific agent taste and preferences.
  • If the task mentions tickets, contextspace, dispatch, or .codex-autorunner/, also read .codex-autorunner/ABOUT_CAR.md.
  • If you need runtime setup details, read .codex-autorunner/DESTINATION_QUICKSTART.md.

Hard Invariants

  • Filesystem state is the source of truth.
  • Durable runtime state lives under .codex-autorunner/.
  • Keep the layered split intact: protocol-agnostic engine, adapters, surfaces.
  • Prefer deterministic behavior, explicit config, small diffs, and observable failures.
  • Agents can propose and execute, but files decide.

Fast Routing

  • Core engine and config: src/codex_autorunner/core/, docs/car_constitution/10_CODEBASE_CONSTITUTION.md, 20_ARCHITECTURE_MAP.md, 30_ENGINEERING_STANDARDS.md.
  • Web UI frontend: src/codex_autorunner/web_frontend/AGENTS.md.
  • Web backend and static asset serving: src/codex_autorunner/surfaces/web/AGENTS.md.
  • Frontend and web tests: tests/AGENTS.md.
  • Agent protocol work: src/codex_autorunner/agents/acp/AGENTS.md.
  • Debugging and ops: docs/car_constitution/50_OBSERVABILITY_OPERATIONS.md, docs/ops/.

CAR Files

  • Update tickets only under .codex-autorunner/tickets/.
  • Update context docs only under .codex-autorunner/contextspace/.
  • Do not create duplicate ticket or context files elsewhere in the repo.
  • Treat .codex-autorunner/ as intentional structure even when generated or gitignored.

Repo Defaults

  • Python: use the project venv, usually via .venv/bin/python; run make setup if needed.
  • Frontend JS: edit src/codex_autorunner/web_frontend/src/; use make serve for Vite HMR against the local hub, or pnpm run build / make build to refresh web_static/.
  • Prefer make targets for common checks.
  • Use gh for GitHub operations when possible.
  • Use a 300 second timeout for git commit; avoid --no-verify unless the user approves.
  • Do not restart launchd services or run refresh scripts yourself; ask the user to do that. Preferred update path is scripts/safe-refresh-local-mac-hub.sh.

Useful Pointers

  • Release flow: docs/ops/release.md
  • Telegram debugging: docs/ops/telegram-debugging.md
  • VM/cloud caveats: docs/ops/running-in-a-vm.md
  • Subagent model configuration: docs/adding-an-agent.md

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.