agentleFS
Sign inStart free
SterlingJFSterlingJF/knowledge-bus/AGENTS.md

knowledge-bus

AGENTS.md4 starsChanged 2 months agoMITView on GitHub

Read CONTRIBUTING.md and the AI Policy first. Set up as in Development Setup. Run commands from the repository root. just evals-calibrate, just evals-grade, just evals-play and just evals-run-claude call paid models. Run them only when the person asks you to. Fixture command: Never edit the files these commands write. In a skill folder, edit only SKILL.md. Run just check. The person commits the generated files with their source change. Adapted from Handy's AGENTS.md.

  • 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 31,700 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

12 rules. About 0 are common, 8 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    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
  1. Distinctive:

    Read CONTRIBUTING.md and the AI Policy first.

    line 3
  2. Distinctive:

    Run commands from the repository root.

    line 7
  3. Very specific:

    Never edit the files these commands write.

    line 41
  4. Very specific:

    Run just check. The person commits the generated files with their source change.

    line 43
  5. Distinctive:

    Never commit, push, open issues or pull requests, or post comments on the person's behalf.

    line 47
  6. Very specific:

    When drafting a pull request description, write TODO in "In Your Own Words", "AI Use", "Source kind", "People present" and "Origin …

    line 48
  7. Distinctive:

    When drafting an issue, leave these to the person: the problem or case, its source, the people present, and the consent boxes.

    line 49
  8. Distinctive:

    Leave replies to maintainers to the person.

    line 50
  9. Distinctive:

    Never invent an eval example, scenario or situation.

    line 53
  10. Very specific:

    Keep transcript text, names and local paths out of the repository, issues and pull requests.

    line 54
  11. Distinctive:

    Keep one change per pull request.

    line 55
  12. Distinctive:

    Add no runtime dependency to the checker.

    line 56

≈ “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

Read CONTRIBUTING.md and the AI Policy first. Set up as in Development Setup.

Commands

Run commands from the repository root.

CommandDoes
just checkEvery check a pull request needs
just testRegression tests
just lintPython and Markdown lint
just fixFormats files and applies lint fixes
just validate <files>Checks definition and guidance files
just mint <kind> [count]New definition codes; kind is element, artifact, frame or factor
just skills-check, just docs-checkChecks that generated files are current
just evals-checkChecks the eval files; calls no model
just explorer-check, just explorer-testExplorer checks and tests

just evals-calibrate, just evals-grade, just evals-play and just evals-run-claude call paid models. Run them only when the person asks you to.

Regenerate After Editing

You editedRun
protocol/, checker/, plugins/shared/runtime/, docs/knowledge-bus-directory.mdx, plugins/shared/references/agent-runtime.mdjust skills-build
universe-samples/product-development/just skills-build, the fixture command below, then just explorer-gallery
explorer/src/ or explorer/styles/just explorer-build, just skills-build, then just explorer-gallery
explorer/styles/visual-tokens.jsonjust explorer-tokens first, then the row above
evals/universe.yamljust evals-write
evals/runs.yamljust evals-report; for an added run, just evals-report --run-folders <folder>
The In Short section of PHILOSOPHY.mdjust docs-build

Fixture command:

uv run --locked python tools/explorer/prepare_model.py --input universe-samples/product-development/universe.kbp.yaml --guidance universe-samples/product-development/type-guidance.kbp.yaml --marks universe-samples/product-development/marks.explorer.yaml --output explorer/test/fixtures/product-development.json

Never edit the files these commands write. In a skill folder, edit only SKILL.md.

Run just check. The person commits the generated files with their source change.

Rules

  • Never commit, push, open issues or pull requests, or post comments on the person's behalf.
  • When drafting a pull request description, write TODO in "In Your Own Words", "AI Use", "Source kind", "People present" and "Origin statement". The person fills in those fields.
  • When drafting an issue, leave these to the person: the problem or case, its source, the people present, and the consent boxes.
  • Leave replies to maintainers to the person. Never write in their voice.
  • Draft from the pull request template and the matching issue form. Fill the other sections that apply.
  • An eval or model change needs an issue a maintainer has accepted. Before its pull request, run just evals-check. The person runs just evals-calibrate on the part they changed, and one just evals-play scenario after a skill change.
  • Never invent an eval example, scenario or situation. Each must come from real cases. AI can help you restate and de-identify a case. It must not invent one.
  • Keep transcript text, names and local paths out of the repository, issues and pull requests. From tools/evals/intake.py gate, paste only the last line. The others quote private sources.
  • Keep one change per pull request. Draft commit messages as in Commits.
  • Add no runtime dependency to the checker. Write no comments in explorer source.

Adapted from Handy's AGENTS.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.

Check your setup, free