agentleFS
theamtheam/claude-dev-kit/CLAUDE.md

claude-dev-kit

CLAUDE.md17 starsChanged 3 months agoApache-2.0View on GitHub

Conventions for anyone (human or agent) working in this repo. This is the repo that is the kit, so we dogfood it: develop new work through the kit's own flow — file it as a GitHub issue, run /fullstack-dev-kit:work-story <#> (plan gate → implement → gates → PR → pr-review/fix-pr), and let the gates below apply to the kit's own PRs. See CONTRIBUTING.md for the full contributor guide; this file is the short, imperative version the agent reads first. A…

The file

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

1 rule. About 0 are common, 1 is distinctive, and 0 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.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Telemetry privacy is non-negotiable: never send prompts, code, file names, ticket contents, emails, or org data (see TELEMETRY.md).

    line 35

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.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
CLAUDE.md

CLAUDE.md — claude-dev-kit

Conventions for anyone (human or agent) working in this repo. This is the repo that is the kit, so we dogfood it: develop new work through the kit's own flow — file it as a GitHub issue, run /fullstack-dev-kit:work-story <#> (plan gate → implement → gates → PR → pr-review/fix-pr), and let the gates below apply to the kit's own PRs. See CONTRIBUTING.md for the full contributor guide; this file is the short, imperative version the agent reads first.

What this repo is

A Claude Code / Agent-Plugins plugin: agents, skills, commands, and instructions written in Markdown, plus a small zero-dependency Node build/validate/test harness. It is intentionally project- and stack-agnostic — nothing client-, company-, or project-specific belongs here, and no skill/agent hardcodes a language, framework, or test runner (that lives in consuming repos). The one exception is the enumerated tracker / PR-host / design adapters.

Canonical sources vs. the generated bundle

  • Edit the sources: agents/, skills/, codex/skills/, commands/, instructions/.
  • Never hand-edit plugins/fullstack-dev-kit/ — it is generated. Change the source, then run the build. scripts/lib/bundle-sources.mjs is the single source of truth for what is copied/derived; the three bundle manifests (.codex-plugin / portable / cursor) are derived from .codex-plugin/plugin.json, and the version is synced from .claude-plugin/plugin.json.
  • The two contract.v1.json copies (telemetry/ + packages/telemetry-relay/) must stay byte-identical.

The gates (this repo's own standard — what work-story/create-pr must run here)

This is a Markdown+scripts repo, so the usual coverage-% / e2e gates do not apply — do not scaffold a coverage tool or an e2e framework here. The gate is the harness below; all of it must pass before a PR, and a rebuild must leave the tree clean:

node scripts/build-codex-plugin.mjs   # refresh plugins/fullstack-dev-kit/ from sources
node scripts/validate-codex-plugin.mjs
node --test scripts/*.test.mjs        # drift guard, contract parity, PR-detect, stack-profile completeness
git diff --exit-code                  # a dirty tree = the bundle wasn't rebuilt/committed

CI (.github/workflows/ci.yml, the verify job) runs exactly this. A gate that genuinely doesn't apply is reported as N/A with a reason, never silently skipped.

Writing conventions

  • Words are the code. Agents and skills are prompts: imperative, short, matching the existing tone. Don't add a rule that another file already enforces.
  • Frontmatter: agents use model: inherit + a skills: list; skills use name + a trigger-worthy description; commands use description (a usage line) + $ARGUMENTS.
  • One concern per PR, small reviewable diffs, conventional-commit titles (fix: / feat: / docs: / feat!:).
  • Don't weaken the gates — plan approval, security pass, and the adaptive quality gates are the product; relaxing them needs a strong case in the PR.
  • Docs travel with behavior: a change to a skill/agent updates the README.md "What's inside" table (and any affected flow) and adds a CHANGELOG.md [Unreleased] entry, in the same PR.
  • Telemetry privacy is non-negotiable: never send prompts, code, file names, ticket contents, emails, or org data (see TELEMETRY.md).

Versioning (read before touching the version)

version in .claude-plugin/plugin.json is what reaches installed clients — a change is only "live" once it's bumped, rebuilt, and released (see CONTRIBUTING.md → Releases). A version bump reaches users only via a rebuild, so bump and build-codex-plugin.mjs go together.

⚠️ Heads-up: setting the version inside a feature PR has caused collisions (two PRs claiming the same number). We're moving version management to a release job on main — see #87. Until that lands, if a change should reach users, bump the patch/minor and rebuild; once #87 lands, stop editing the version file in PRs and add a changeset instead. This file is updated as part of that migration.

Config that is not bundled into the plugin (this file, .claude/dev-kit.json, docs) needs no version bump.

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.