4. Configuring Your First Project

Why configuring your project matters

Every chapter so far has been theory and setup. Mental models, architecture, installation.

In this Chapter, you will configure your Claude with your project use case. You will choose a real project, configure Claude to work on it autonomously, and commit the infrastructure as a configuration to your project repo that makes every future session faster.

The project you pick here is the project you build for the rest of this course. Not yet another to-do app or a simple prototype, pick the thing you actually want to ship.

Before I built a granular permission system, during a half an hour Claude Code session building a new API route, I was interrupted 16 times with permission prompts.

• "Do you want to allow Bash(npm run test)?" Yes.
• "Do you want to allow Bash(npm run typecheck)?" Yes.
• "Do you want to allow Edit(src/routes/api.ts)?" Yes.

Each interruption took 2 to 3 seconds to read and approve. That adds up to over a minute of pure friction, not counting the context switches that broke my concentration.

By the end of this chapter, you will understand iterating towards a permission configuration that lets Claude work without interruptions, while still blocking force pushes, recursive deletions, and production database commands.

You will walk away with five outcomes:

1. A project you chose (from a custom project idea generator skill or an existing codebase)
2. The .claude/ directory configured as your agent's operating system
3. A settings.json with a complete permissions profile
4. Claude config defaults set for productive daily use
5. A self-evaluating hook for the permission requests