agentleFS
Sign inStart free
PaldomPaldom/playwright-stealth/AGENTS.md

playwright-stealth

AGENTS.md0 starsChanged 2 months agoMITView on GitHub

Canonical agent instructions for this repository — CLAUDE.md simply imports this file, so every agent (Claude Code, Copilot, Cursor, Codex, …) reads the same rules. Agent Skills repository — skills live in skills/<name>/ (one purpose per skill), distributed via the plugin manifest in .claude-plugin/. - Validate everything (the only gate): make check — validator (frontmatter, evals, security rules, manifests), make lint (ruff), make evals (trigger scoring), then the self-checks - Install the commit-time hooks once per clone: make hooks -…

  • Commits and pushes

15 other projects use a version of this.

Adopted by

Most recently taken up by

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

    Eval-first: write skills/<name>/evals/evals.json before the SKILL.md body.

    line 19
  2. Very specific:

    Every added/changed skill updates the README catalog table, CHANGELOG.md, and its skills.sh.json grouping.

    line 31
  3. Very specific:

    Never run git commit or git push: Leave every change in the working tree for the owner to review and commit.

    line 36
  4. Distinctive:

    ALL of it recursively, never commit its contents, never cite it as a committed path.

    line 41

≈ “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

Canonical agent instructions for this repository — CLAUDE.md simply imports this file, so every agent (Claude Code, Copilot, Cursor, Codex, …) reads the same rules.

Agent Skills repository — skills live in skills/<name>/ (one purpose per skill), distributed via the plugin manifest in .claude-plugin/.

Commands

  • Validate everything (the only gate): make check — validator (frontmatter, evals, security rules, manifests), make lint (ruff), make evals (trigger scoring), then the self-checks
  • Install the commit-time hooks once per clone: make hooks
  • Validate one file: python3 scripts/validate_skills.py --file skills/<name>/SKILL.md

Non-negotiable conventions

  • Eval-first: write skills/<name>/evals/evals.json before the SKILL.md body.
  • Frontmatter description is a single line (multi-line silently disables the skill), third person. Model-invoked skills add trigger phrasings and a "Not for …" exclusion; user-invoked ones (disable-model-invocation: true) are never routed, so they get one verb-first line and no trigger cases.
  • name equals the folder name, kebab-case.
  • Shipped Python is stdlib-only; ruff and pre-commit are dev tooling pinned in requirements-dev.txt. Fix a ruff finding at the line with a reason, never by dropping the rule. Never suppress a security finding to get green.
  • SKILL.md bodies target 25–150 lines (< 500 is the hard ceiling); long material goes to references/ (linked one level deep); deterministic steps go to scripts/ with non-zero exit on failure.
  • Every added/changed skill updates the README catalog table, CHANGELOG.md, and its skills.sh.json grouping.
  • Publication to skills.sh happens ONLY via the bundled /publish-repo skill (slash-invoked, needs the owner's go-ahead) — never flip repo visibility ad hoc. Deployment model: docs/deploying.md.
  • Never run git commit or git push. Leave every change in the working tree for the owner to review and commit. (The bash-guard hook additionally blocks --no-verify and force-pushes as a safety net; the server-side main ruleset is the real gate.)
  • .local/ is gitignored personal material (only its README is committed) — read ALL of it recursively, never commit its contents, never cite it as a committed path.

Where things are

  • Authoring rules: docs/skill-authoring.md · Eval methodology: docs/evals.md
  • Session goal prompt: .local/PROMPT.md — personal, created at scaffold time, never committed; if it is missing, ask the maintainer for their goal prompt.
  • Hooks: .claude/hooks/ (SKILL.md write-time validation, bash guard) — wired in .claude/settings.json; changes to them get PR-level scrutiny.
  • The add-skill skill (.claude/skills/add-skill/) walks the authoring workflow — prefer it over ad-hoc skill writing. The publish-repo skill (.claude/skills/publish-repo/) walks skills.sh deployment.

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