agentleFS
Sign inStart free
DegreeSignDegreeSign/skills/AGENTS.md

skills

AGENTS.md1 starsChanged 18 days agoMITView on GitHub

  • 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 repository8 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

21 rules. About 0 are common, 11 are distinctive, and 10 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    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. Distinctive:

    Use neutral placeholders for paths, hosts, environment variable names, service names and deploy commands.

    line 6
  2. Distinctive:

    Always use a fresh generic name.

    line 7
  3. Distinctive:

    Every rule must be generic, state a required behaviour and never a preference.

    line 11
  4. Distinctive:

    Never put specific values or examples in a rule.

    line 12
  5. Very specific:

    Create one folder per sub-skill at skills/<name>/.

    line 17
  6. Very specific:

    Every sub-skill must have SKILL.md with name and description frontmatter.

    line 18
  7. Very specific:

    Put supporting files beside SKILL.md and reference them with relative paths.

    line 19
  8. Distinctive:

    Keep each sub-skill self-contained.

    line 20
  9. Very specific:

    Add the new sub-skill to the index in the root SKILL.md.

    line 21
  10. Very specific:

    Write skills generically. Never mention the projects, repos or packages that inspired a skill.

    line 25
  11. Distinctive:

    Keep every skill project-agnostic.

    line 26
  12. Very specific:

    Keep the frontmatter description focused on the skill's most important aspects.

    line 28
  13. Distinctive:

    When the scope is unclear, ask the user with concrete options before writing.

    line 29
  14. Very specific:

    After changing a skill, refresh the README.md skills table and ChangeLog.md.

    line 34
  15. Very specific:

    Keep the skills badge count in README.md equal to the number of folders under skills/.

    line 35
  16. Very specific:

    Verify installer changes with yarn build, then node bin/install.mjs --list and node bin/install.mjs --dry-run.

    line 41
  17. Distinctive:

    Never publish, tag, stage or commit the package.

    line 42
  18. Distinctive:

    Ask the user before applying an update.

    line 47
  19. Very specific:

    Keep ChangeLog.md to public changes only, one line per change.

    line 52
  20. Distinctive:

    Never commit or stage changes.

    line 58
  21. Distinctive:

    Never unstage changes either.

    line 59

≈ “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

General

  • Apply these rules to every skill in this repository.
  • Use neutral placeholders for paths, hosts, environment variable names, service names and deploy commands.
  • Always use a fresh generic name; never reuse a specific one, and generalise the whole flow, not just the names.

Adding a rule

  • Every rule must be generic, state a required behaviour and never a preference.
  • Never put specific values or examples in a rule.

Adding a skill

  • The root SKILL.md is the entry skill. It indexes the sub-skills and is the only skill that references them.
  • Create one folder per sub-skill at skills/<name>/.
  • Every sub-skill must have SKILL.md with name and description frontmatter. name must match the folder and use lowercase hyphen-separated words.
  • Put supporting files beside SKILL.md and reference them with relative paths.
  • Keep each sub-skill self-contained. A sub-skill must not read files from another sub-skill.
  • Add the new sub-skill to the index in the root SKILL.md.

Writing a skill

  • Write skills generically. Never mention the projects, repos or packages that inspired a skill; use neutral example names such as mylib.
  • Keep every skill project-agnostic. A skill must apply to any project and must not assume a particular layout, language, tool, path or domain.
  • Research the real source material first, then generalise. Do not invent behaviour or APIs.
  • Keep the frontmatter description focused on the skill's most important aspects. Do not use colons.
  • When the scope is unclear, ask the user with concrete options before writing.

Changing a skill

  • Adding a skill is a minor version bump.
  • After changing a skill, refresh the README.md skills table and ChangeLog.md.
  • Keep the skills badge count in README.md equal to the number of folders under skills/.

Installer

  • The installer source is src/install.mts; the bundled bin/install.mjs is generated by yarn build.
  • The installer is bin/install.mjs; the user runs it, never the agent.
  • Verify installer changes with yarn build, then node bin/install.mjs --list and node bin/install.mjs --dry-run.
  • Never publish, tag, stage or commit the package.

Updates

  • The root SKILL.md documents a daily update check. Run check.mjs at most once per machine per day; it caches the timestamp in the XDG cache directory.
  • Ask the user before applying an update. Never update on your own.

Version and changelog

  • Bump VERSION and add a ChangeLog.md entry for every release.
  • Keep ChangeLog.md to public changes only, one line per change.

Commit messages

  • Format a release commit as version followed by the release version, with no other words.
  • Name every other commit with 2 to 4 lowercase words, letters only.
  • Never commit or stage changes. Only the user commits or stages, unless the user gives an explicit order.
  • Never unstage changes either. Do not run git reset, git restore --staged, git rm --cached, or any other command that alters the index. The user alone controls the index.

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