Claude Code Setup Prompt

# Claude Code Setup Prompt

Copy and paste this in Claude Code at the root of your project:

---

```
I need you to analyze this project and generate the complete infrastructure to work efficiently with Claude Code following best practices. Do the following:

## 1. CLAUDE.md (project root)

Analyze the code, structure, dependencies, and patterns of this project. Generate a CLAUDE.md that includes:

- Project description and business purpose
- Tech stack and main versions
- Directory structure with explanation of each important folder
- Code conventions (naming, patterns, architecture)
- Essential commands (build, test, lint, deploy)
- Team workflow (branches, PRs, CI/CD)
- Common mistakes to avoid (based on what you see in the code)
- Relevant business context to understand technical decisions

## 2. .claude/commands/

Create slash commands for common workflows. At minimum these files:

- `pr.md` - Create a PR with description, tests, and changelog
- `review.md` - Review code looking for bugs, security issues, and improvements
- `test.md` - Generate tests for a specific file or function
- `refactor.md` - Refactor code while maintaining functionality
- `docs.md` - Generate/update documentation
- `debug.md` - Investigate and fix a bug
- `feature.md` - Implement a feature from plan to PR

Each command should include clear instructions, expected output format, and verification steps.

## 3. .claude/settings.json

Configure safe pre-approved permissions based on the project's tools:

- Read commands (cat, ls, find, grep, etc.)
- Detected package manager commands (npm, yarn, pnpm, pip, etc.)
- Safe git commands (status, diff, log, branch)
- Project linters and formatters
- Test commands
- Any specific CLI you detect (kubectl, docker, mongosh, etc.)

DO NOT include destructive commands or those with dangerous side-effects.

## 4. .claude/agents/

Create specialized subagents:

- `code-simplifier.md` - Simplify and clean code without changing functionality
- `security-check.md` - Security audit of the code
- `verify.md` - Verify that changes work correctly

## 5. .claude/hooks/ (documentation)

Create a `hooks-setup.md` file explaining how to configure:

- PostToolUse hook for automatic formatting with the project's formatter
- Stop hook for automatic verification on long-running tasks

## 6. .mcp.json (if applicable)

If you detect integrations that could benefit from MCP servers (Slack, databases, APIs), create a template .mcp.json with commented structure.

## 7. .claude/ README

Create `.claude/README.md` explaining the entire generated structure and how the team can contribute/modify it.

---

IMPORTANT:
- Adapt everything to the REAL context of this project, don't use generic placeholders
- Use the style and conventions that already exist in the code
- If you find existing documentation, integrate it
- Verify that the commands you pre-approve actually exist in the project
- Include concrete examples based on the actual code

Start by analyzing the project and then generate each file. Show the plan before executing.
```

---

## Usage Tips

1. **Run this in Plan mode first** (shift+tab x2) to review what it will do
2. **After generation**, review the CLAUDE.md and adjust what's not right
3. **Commit everything** so the team has it
4. **Iterate**: every time Claude does something wrong, add it to CLAUDE.md

## Quick Command for Later

Once you have the setup, you can add this alias to your shell:

```bash
alias cc="claude --profile work"
```

And create the profile at `~/.claude/profiles/work.json` with specific settings for your org.

Pro tip: add prompt as an skill