agentleFS
Sign inStart free
gastonlarap-a11ygastonlarap-a11y/claude-code-setup/AGENTS.md

claude-code-setup

AGENTS.md1 starsChanged 3 months agoMITView on GitHub

Portable Claude Code configuration, the dev-plugins marketplace, and a protocol that configures AI coding agents in any project (Claude Code natively; Codex/Cursor read the generated AGENTS.md directly, and so does Antigravity CLI). Agent entry points: - Not sure / any machine, any user: START.md — interactive bootstrap that detects the environment, interviews the user, and routes to the right procedure below. - Restore a machine (owner only): AGENT-INSTALL.md → runs install.sh, which OVERWRITES ~/.claude/. Never suggest it to configure a…

  • 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 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 repository4 other files it gives its agents.

The file

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

3 rules. About 0 are common, 1 is distinctive, and 2 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.

  • 2

    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:

    All config content in English.

    line 30
  2. Very specific:

    Before declaring work done, run the CI checks locally and report real results — the verify skill runs the full validate.yml mirror (check …

    line 42
  3. Distinctive:

    No speculative abstractions: a new installer path, script or config knob needs a problem this repo already has — say which and why.

    line 47

≈ “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

claude-code-setup

Portable Claude Code configuration, the dev-plugins marketplace, and a protocol that configures AI coding agents in any project (Claude Code natively; Codex/Cursor read the generated AGENTS.md directly, and so does Antigravity CLI). Agent entry points:

  • Not sure / any machine, any user: START.md — interactive bootstrap that detects the environment, interviews the user, and routes to the right procedure below.
  • Restore a machine (owner only): AGENT-INSTALL.md → runs install.sh, which OVERWRITES ~/.claude/. Never suggest it to configure a mere project or for other users.
  • Configure/improve any project (anyone): AGENT-PROJECT-SETUP.md → executes global/skills/setup-project/SKILL.md.
  • Refresh curated knowledge: /refresh-knowledge (recipes/references vs current docs).

Editing rules for this repo

  • Plugins are edited HERE under plugins/, never in ~/.claude/. Any plugin change bumps version in BOTH .claude-plugin/marketplace.json (that entry) and the plugin's own .claude-plugin/plugin.json, then validate & republish: claude plugin validate ./plugins/<name> --strict && claude plugin marketplace update dev-plugins && claude plugin update <name>@dev-plugins. CI re-checks JSON, version parity + enabledPlugins sync, forgotten version bumps (feature branches), schemas, language anchors, PowerShell lint, template-skeleton validity and secret scanning (gitleaks) on push (validate.yml); installer changes also trigger the real-execution smoke matrix (installer-smoke.yml, Linux + Windows PowerShell 5.1).
  • Global config lives in global/ and reaches ~/.claude/ only via install.sh.
  • Token discipline: skill bodies and references/ load on demand — keep descriptions sharp and short; CLAUDE.md-style content stays within the budgets defined in global/skills/setup-project/SKILL.md.
  • All config content in English; README stays in Spanish. secrets.env is never committed.
  • Secrets: secrets.env only feeds the installer's user-scope MCP registrations (global/mcp-servers.json, data-driven). A future plugin-scoped secret uses userConfig with sensitive: true in its plugin.json (stored in the OS keychain, prompted at enable time) — never a new bespoke installer path. Hooks consume it via exec form + the CLAUDE_PLUGIN_OPTION_<KEY> env var; ${user_config.*} is rejected in shell-form hook commands (2.1.207 security fix).
  • Repo-level changes (installers, global/, CI, protocols) get a CHANGELOG.md entry under [Unreleased]; per-plugin changes rely on the dual-bump rule instead.

Engineering standards

  • Before declaring work done, run the CI checks locally and report real results — the verify skill runs the full validate.yml mirror (check scripts, shellcheck, plugin schema validation).
  • Error contract: hooks and statusline scripts are fail-open (degrade silently, never break a session); installers and CI scripts fail loud. Keep new scripts on the right side.
  • No speculative abstractions: a new installer path, script or config knob needs a problem this repo already has — say which and why.
  • Ambiguous request → ask targeted questions first; a better route exists → say why and let the requester choose before proceeding.

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