agentleFS
seancabalseseancabalse/agentic-playbook/skills/bootstrap-project/SKILL.md

bootstrap-project

Skill0 starsChanged 15 days agoMITView on GitHub

Set up the playbook in a new or existing repository: AGENTS.md, ADR 0001, the living docs, CI and owner tasks. Run only when the user asks.

13 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 12 edited versions we read keep these.

  • 3

    Distinctive

    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
  1. Very specific:

    Read the repo (README, package manifests, existing docs, CI) or, for a new project, the brief or PRD.

    line 14
  2. Very specific:

    Run the installer from the plugin: "${CLAUDE_PLUGIN_ROOT}/scripts/playbook.sh" project <repo> --stack <stacks> --tracker …

    line 21
  3. Very specific:

    Add cursor or copilot to --tools only if the team uses those tools.

    line 22
  4. Distinctive:

    Keep sections 1–4 (Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution) as they are.

    line 29
  5. Distinctive:

    Keep the chosen stack overlays.

    line 35
  6. Distinctive:

    If an AGENTS.md already existed, merge into it: the project's rules win over the playbook defaults.

    line 36

“Common” is measured: kept by at least 6 in 10 of the 12 edited versions of this file we could read. “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
SKILL.md

Bootstrap a project

Stacks: $ARGUMENTS

1. Understand the project

  • Read the repo (README, package manifests, existing docs, CI) or, for a new project, the brief or PRD. Infer the stacks from the manifests ("Stack inference" at the end of this file), and list the stack files that exist: ${CLAUDE_PLUGIN_ROOT}/stacks/.
  • Ask the owner, in one batch, only what you can't infer: what it is and for whom, the tracker (Linear, Jira, GitHub Issues, or another one) and how agents reach it (MCP server, CLI), the tenant unit, the target platforms, hosting preferences, and any compliance needs.

2. Install the playbook files

You're running from the installed plugin, so the project needs no copy of the playbook for Claude Code.

  1. Run the installer from the plugin: "${CLAUDE_PLUGIN_ROOT}/scripts/playbook.sh" project <repo> --stack <stacks> --tracker <linear|jira|github>. The default --tools plugin,agents builds AGENTS.md from the playbook plus the stacks and the tracker section, writes CLAUDE.md, and sets worktree.baseRef in .claude/settings.json.
  2. Add cursor or copilot to --tools only if the team uses those tools. They read files in the repo, so the installer then also copies the playbook into .playbook/ (committed, and never a plugin source).
  3. Never register a project's .playbook/ as a plugin (claude plugin marketplace add ./.playbook, or extraKnownMarketplaces pointing at it). Plugin sources are per machine: it would replace the machine-wide install for every project. If the repo has one from an older install, the installer warns: remove it, and the copy too unless Cursor or Copilot needs it.

Project changes to a playbook file go in .playbook/local/<same path> (the customize skill). For a tracker with no file in trackers/, write the "Tracker" section by hand from the GitHub one, covering access, ids and linking, the status mapping, where the docs live, and the editing notes.

3. Fill AGENTS.md

  • Keep sections 1–4 (Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution) as they are. Adapt sections 5–7 only where the project needs to, and tell the owner what changed.
  • Fill section 8, "This project":
    • What it is, and the tracker with doc links.
    • Stack and docs URLs for the installed versions.
    • Commands, verified by running each one.
    • Layout, environments, and the tenant unit.
  • Keep the chosen stack overlays; delete the lines that don't apply, and add project-specific traps.
  • If an AGENTS.md already existed, merge into it: the project's rules win over the playbook defaults; note conflicts for the owner.
  • CLAUDE.md contains @AGENTS.md.

4. Decisions and docs

  • ADR 0001: stack (write-adr), with the options considered. More ADRs for the other hard choices already made (auth, data, hosting).
  • Tracker docs, from ${CLAUDE_PLUGIN_ROOT}/templates/:
    • Decision Log (seeded with the bootstrap decisions).
    • Architecture & Data Model.
    • Developer Runbook.
    • PRD (from the brief, with open questions listed).
  • Link them from AGENTS.md.

5. Safety rails

  • Worktrees: .gitignore has .claude/worktrees/; .claude/settings.json sets "worktree": { "baseRef": "head" } so agents' worktrees start from the current branch; .worktreeinclude (from ${CLAUDE_PLUGIN_ROOT}/templates/worktreeinclude) lists the gitignored files a worktree needs to run the app and tests.
  • .gitignore covers env files, generated folders, build output, local design files and keys.
  • CI runs lint, typecheck, unit tests and the dependency audit (plus DB and integration tests if there's a backend), with pinned tool versions.
  • Secrets plan: the names needed per environment, and the owner sets each value (give the exact commands; never handle the values).
  • Branch protection on the default branch (owner task if you lack permission).
  • Squash merging only, with merged branches deleted (principles/10-delivery-ci-release.md #3; owner task if you lack permission): gh repo edit --enable-squash-merge --enable-merge-commit=false --enable-rebase-merge=false --squash-merge-commit-message pr-title-description --delete-branch-on-merge.

6. Owner tasks

File tickets with step-by-step instructions for accounts, billing, DNS and credentials, labelled as owner tasks.

7. Report

What's set up, what's verified (the commands ran green), the open questions, and the owner tasks.

Stack inference

Suggest these stacks from what the repo contains, then confirm them with the owner:

  • nx.json → nx-monorepo.
  • package.json with React for the browser (react-dom, Next.js, Remix or React Router, Vite with React) → react-web.
  • package.json with a Node server (Express, Fastify, NestJS, Hono) → node-service.

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.