Claude Code Project Setup Checklist
First-hour setup for a Claude Code project: CLAUDE.md, permissions, first skills, and hooks, in order. Copy it below, fill the bracketed fields, delete what you don't need.
# Claude Code setup: [project]
## 1. CLAUDE.md (15 min)
- [ ] Project context: two sentences
- [ ] Commands: test, build, run, with ports (verified by running them)
- [ ] Code style rules linters don't cover
- [ ] Gotchas section started (even if empty)
## 2. Permissions (10 min)
- [ ] Review default permission mode for this repo's risk level
- [ ] Allowlist routine safe commands (test, lint, build)
- [ ] Confirm destructive operations still prompt
## 3. First skills (20 min)
- [ ] One process skill: verification or review (superpowers or equivalent)
- [ ] One project skill: your deploy/test workflow as .claude/skills/ or commands
- [ ] Trigger-test each with 3 paraphrased prompts
## 4. Hooks (optional, 15 min)
- [ ] Audit log hook if governance requires it
- [ ] Session-start context injection if you maintain memory files
## 5. Verify the setup
- [ ] Cold-start test: fresh session lands a trivial change using only the repo
- [ ] The wrong-answer test: ask it the port/command you documented; it should know
Field notes
| Field | Why it matters |
|---|---|
| Verified by running them | Agents trust CLAUDE.md commands completely. An unverified command is a planted bug with your name on it. |
| Cold-start test | The acceptance test for the whole setup: a context-free session should succeed using only what you wrote down. |