agentleFS
ANT1OFFANT1OFF/gram/AGENTS.md

gram

AGENTS.md1 starsChanged yesterdayMITView on GitHub

Gram is a desktop app for individual developers who run their own installed Claude Code and Codex CLIs on their own subscriptions. README introduces it, contributing covers branches, checks and pull requests, and the docs map leads to the rest. - Keep gateways opt-in, subscription credentials isolated, helper context consent separate per provider and scope, and approval and ownership boundaries intact. - Keep personal infrastructure, private hosts and repositories, recorded private data and corporate defaults out of the tree. Fixtures…

  • Reads credentials
  • 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 32,671 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

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

8 rules. About 1 is common, 3 are distinctive, and 4 are very specific to this project.

  • ≈1

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Common:

    Keep gateways opt-in, subscription credentials isolated, helper context consent separate per provider and scope, and approval and ownership …

    line 8
  2. Distinctive:

    Keep personal infrastructure, private hosts and repositories, recorded private data and corporate defaults out of the tree.

    line 10
  3. Very specific:

    Never read .env files, key files or credential stores, and never put secrets, environment dumps or boot URLs in logs, commits or PRs.

    line 13
  4. Very specific:

    Run bun run check:scaffold and the relevant checks from the verification matrix.

    line 17
  5. Very specific:

    Never stop processes by pattern (pkill -f, killall, or kill on pgrep matches): other checkouts may share the machine.

    line 20
  6. Distinctive:

    Don't test wording (prompts, skills, tool descriptions, UI copy), mirrors of the implementation or the same thing at two layers.

    line 27
  7. Distinctive:

    Use 24-hour times in product UI and documentation.

    line 31
  8. Very specific:

    Never push directly to main, including force pushes.

    line 38

≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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

Working on Gram

Gram is a desktop app for individual developers who run their own installed Claude Code and Codex CLIs on their own subscriptions. README introduces it, contributing covers branches, checks and pull requests, and the docs map leads to the rest.

  • Keep gateways opt-in, subscription credentials isolated, helper context consent separate per provider and scope, and approval and ownership boundaries intact.
  • Keep personal infrastructure, private hosts and repositories, recorded private data and corporate defaults out of the tree. Fixtures are synthetic and set provider, auth and CLI state explicitly; they never inherit the host's GitHub login or a paid account.
  • Never read .env files, key files or credential stores, and never put secrets, environment dumps or boot URLs in logs, commits or PRs.
  • Preserve exact dependency pins, the workspace catalog, native overrides and the Effect patch unless the task needs a change.
  • Run bun run check:scaffold and the relevant checks from the verification matrix. Focused checks while iterating; bun run check is the full integration check.
  • Never stop processes by pattern (pkill -f, killall, or kill on pgrep matches): other checkouts may share the machine. Stop your own check with bun run check:cancel, or kill only the exact PIDs your command started.
  • Fix demonstrated causes, with a before/after reproduction where feasible. Never weaken assertions, skips, timeouts or permission guards to get a pass. Claim only what was tested: synthetic success is not live-provider qualification.
  • A test earns its place by guarding behaviour that could break and that someone would mind. Don't test wording (prompts, skills, tool descriptions, UI copy), mirrors of the implementation or the same thing at two layers; never cut security, approvals, egress, migrations, data loss, the API/MCP contract or the merge and release paths. See what earns a test.
  • Use 24-hour times in product UI and documentation.
  • For a UI change whose look is open, follow DESIGN.md and the mock-first playbook; most tasks skip the mock.
  • Gram's web app is also mounted inside other apps, which ship a pinned commit of it. When a change removes, renames or re-words something they can see (a screen, a mount option, a kit control, a theme name, a data-testid), follow keeping embedders in step.
  • Never push directly to main, including force pushes. Push a feature branch and open a PR with a conventional title (feat:, fix:, perf:, refactor:, docs:, chore:). Put commands, results and known limits in the PR description.
  • Risky, cross-cutting or security-sensitive changes get an independent review of the current head before they merge; see contributing.

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.