agentleFS
Sign inStart free
relai-airelai-ai/agent-eval-loop/AGENTS.md

agent-eval-loop

AGENTS.md34 starsChanged 3 days agoApache-2.0View on GitHub

Before changing this plugin, read DESIGN_PRINCIPLES.md and follow it. For the current architecture and its rationale, see DESIGN.md. The plugin runs inside whatever coding agent, toolchain and machine the user has. Guidance and shipped code must not assume one of them (principles 11 and 12). - Any coding agent. Do not write agent-specific invocation syntax, such as /plugin:skill slash commands. Refer to a skill by name ("the eval skill"); users start skills with a plain-language request. - No pinned tool…

  • 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 repository8 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, 1 is distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    Before changing this plugin, read DESIGN_PRINCIPLES.md and follow it.

    line 3
  2. Very specific:

    Any coding agent: Do not write agent-specific invocation syntax, such as /plugin:skill slash commands.

    line 11
  3. Distinctive:

    Any container engine the tool supports: Do not require Docker itself.

    line 19
  4. Very specific:

    No fixed architecture: Do not pin a CPU architecture or --platform.

    line 22

≈ “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

Agent instructions

Before changing this plugin, read DESIGN_PRINCIPLES.md and follow it. For the current architecture and its rationale, see DESIGN.md.

Stay portable across users' setups

The plugin runs inside whatever coding agent, toolchain and machine the user has. Guidance and shipped code must not assume one of them (principles 11 and 12).

  • Any coding agent. Do not write agent-specific invocation syntax, such as /plugin:skill slash commands. Refer to a skill by name ("the eval skill"); users start skills with a plain-language request.
  • No pinned tool versions in the plugin. The plugin never names a version of a tool it runs, such as Harbor. At setup, the coding agent records the latest stable release in the user's project ([harbor].version in relai.toml), and every run and checker uses that. Upgrading is a recorded measurement change, never a silent float. Tests of the plugin itself may use a fixed fixture version.
  • Any container engine the tool supports. Do not require Docker itself. Write for any engine with a Docker-compatible CLI (Docker Desktop, OrbStack, Colima, Rancher Desktop) or Podman, as Harbor supports. Checkers detect a running engine rather than assume one.
  • No fixed architecture. Do not pin a CPU architecture or --platform. The machine that runs a test, local or cloud, decides it.
  • Version-specific notes say which release they were checked against. When a note describes a tool's behavior, such as skills/eval/references/harbor-notes.md, name the release it was observed in, and re-check it before relying on it with a newer release.

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