Four files. Five of the fifteen minutes are one step, and skipping it is the difference between a setup and a template.
Last updated: 2026-09-21 · checked against Claude Code 2.1.278
Four files, fifteen minutes, and one of those steps is the whole job. The other three are typing.
| Min | Step | Skippable? |
|---|---|---|
| 0–2 | npm install -g @anthropic-ai/claude-code, then run claude once in your repo to sign in | No |
| 2–5 | Copy CLAUDE.md, .claude/ and .mcp.json.example into your repo root | No |
| 5–10 | Fill in your four commands and your never-do list in CLAUDE.md | This is the step. Skip it and you have a template, not a setup |
| 10–12 | Edit permissions.allow to match your real test and lint commands | Yes, but the prompts continue |
| 12–15 | Run claude, then your ship-check skill. It should run tests and lint and report PASS/FAIL | No — this is the only proof the rest worked |
Claude Code starts every session knowing nothing about your repo. CLAUDE.md is what it reads first, every time. Most people write an essay here; the essay is what makes it useless.
Four commands and a short list of things not to do beats three paragraphs of architecture:
# My Project
## Commands
- install: `pnpm install`
- dev: `pnpm dev`
- test: `pnpm test`
- lint: `pnpm lint --fix`
## Layout
- `src/` app code · `src/db/` migrations, hand-written · `e2e/` Playwright
## Never
- Never edit anything in `src/generated/`
- Never commit to `main`; branch first
- Migrations are forward-only. No editing an applied one
If you only ever do one thing from this page, write the four commands. Everything downstream — the allowlist, the skills, the checks — is about commands Claude already knows how to run.
Approving a command approves that run. It is not remembered, which is why you can approve npm test forty times and be asked a forty-first. The allowlist is what makes it stop:
{
"permissions": {
"allow": [
"Bash(git status)",
"Bash(git diff:*)",
"Bash(npm test:*)",
"Bash(npm run lint:*)"
],
"deny": [
"Bash(rm -rf:*)",
"Bash(git push --force:*)",
"Read(./.env*)"
]
}
}
The detail that catches everyone: Bash(npm test) matches that exact string, so npm test -- --watch still prompts. Bash(npm test:*) covers the arguments too.
Do not allow everything. The point of an allowlist is that you can stop reading the prompts, and that is only safe while the prompts you would want to read still appear. Allow the boring repeated read-only things. Leave deletes, force-pushes and anything that reads .env out.
More detail, including the three reasons a rule silently does not match: Claude Code keeps asking permission.
Four words get treated as one thing. They are not, and three of them can wait:
| Thing | Day one? | Why |
|---|---|---|
| A skill | Yes, one | It is a SKILL.md and it is also the slash command. Start with a ship-check |
| Hooks | Not yet | Useful once you know what should be enforced rather than requested. They also fail quietly |
| An MCP server | Only if needed | Only for reaching something outside your machine. Words cannot query your database |
| A plugin | No | Plugins are about handing a setup to other people. Nothing to package on day one |
Which is which, and the one question that picks between each pair: plugin, skill, subagent, hook or MCP server.
Claude Code watches the settings files, so an edit to permissions or hooks is picked up without restarting the session. If a change seems not to have landed, the configuration is usually wrong rather than stale. claude --debug hooks will tell you which.
Hook stderr goes to the debug log and never to the transcript, so a hook that crashes on line one looks exactly like a hook that was never wired up. And it takes exit 2 to block a tool call — exit 1 is a non-blocking error, so the Unix reflex produces a hook that complains and then allows the thing anyway. Full list: hooks are configured but never run.
All four files, filled in, MIT, nothing held back — kit-claude-code-starter. Clone it, copy three paths, and you are at minute 5 with the only step that matters left to do.
git clone https://github.com/sdvsignal/kit-claude-code-starter
cp -r kit-claude-code-starter/{CLAUDE.md,.claude,.mcp.json.example} your-repo/
cd your-repo && claude
Free starter repo
There is no paid tier of that repo and nothing is withheld from it. If you already know your build commands, this page plus that clone is the whole job. Fifteen minutes gets it installed; proving it actually fires is a separate sitting.
Flat prices, written scope, stop at any rung. No call to book, no retainer.
| Rung | What you get | Clock |
|---|---|---|
| Free — starter repo | A real setup you copy yourself | now |
| $29 — Setup Lite | One project: CLAUDE.md, allowlist, one skill, as a pull request you review | 24 hours |
| $99 — Setup Sprint | 3–5 skills, hooks, up to 3 MCP servers, plus a runbook | 48 hours |
| $79 — Surface Audit | Already set up and still fighting it: contradictions and an exact fix list | 48 hours |
Hooks and MCP servers are in Sprint, not Lite — which one you actually need. Clocks run from a complete intake, not from the Stripe receipt.