📁 Anatomy of a project

New here? Folder (also called a “directory”) is the space on your computer where files are stored together — the same yellow icon you already click in File Explorer or Finder. A project folder it’s a regular folder dedicated to a single project—nothing magical, just organization. The agent can only see what’s inside the folder you opened with claude; mixing several projects in one folder confuses its context, because it starts reading files from different automations as if they were part of the same conversation.

Think of the project folder as a physical file drawer: each drawer holds papers about one subject. If you toss contracts, recipes, and photos into the same drawer, even finding what you need becomes work—the same thing happens when the agent tries to understand a disorganized project. One automation per folder is the rule that prevents this kind of confusion from day one.

New here? One repository (or “repo”) is a project folder with a version history stored by a tool called git. Each significant change is recorded with a date and description — like an automatic "save as" that never deletes the previous version.

You don’t need a repository from your very first project—many simple automations at the beginning of the course run in a regular folder, without git. But once the project becomes valuable (an automation you use every week, for example), it’s worth turning the folder into a repository: if a file breaks or a command accidentally deletes something, you can go back to an earlier version saved in the history.

When working on a project, certain folder and file names come up all the time—recognizing them helps avoid confusion. New here? File is any individual document inside a folder (text, image, code); folders can contain other folders, forming a "tree." A name that starts with a dot, like .claude or .git, it’s a folder hidden — the operating system hides it by default because it's a configuration area, not day-to-day content, but it's always there.

Caption: the project root folder contains CLAUDE.md, the `.claude/` folder with skills, and the code folders — each with a fixed role that repeats in every project in the course.

This structure is repeated in nearly every project you’ll build in the course, so it’s worth memorizing the four names. CLAUDE.md is the only file the agent always reads at the start of every session (module 0.4 explains this in depth); the folder .claude/ is where the "skills" live — processes taught to the agent, the topic of Track 3; and the code folders vary from project to project, but the name src/ (from "source") is the most common.

The project files stay on your computer the whole time. Only the conversation text — and file excerpts the agent needs to read to respond — travel to the cloud-based AI model to generate each response. Nothing gets “published” or made visible to other people just because you talked about it.

Caption: files never leave your computer on their own — only the conversation text and the file excerpts needed to answer travel to the model with each question.

See inside: when you ask it to read a file, Claude Code sends that file’s contents (or the relevant part) along with your message for the AI model to process—that’s how it “sees” what’s in the folder. This has a cost: the more text that gets sent, the more expensive usage becomes (if you’re on a pay-per-use plan) and the longer the response takes. That’s why organized projects, without huge or irrelevant files in the folder, make the agent faster and cheaper to use.

This is the ritual you’ll repeat for every new automation in the course: create the folder, go into it, open the agent. New here? Terminal it’s the text window where you type commands instead of clicking icons—on Windows, it’s usually called “Command Prompt” or “PowerShell”; on Mac and Linux, “Terminal.” Command it’s the instruction you type and confirm with Enter; the computer runs it right away.

Objective: have a new, empty project, ready to receive the CLAUDE.md from module 0.4. How to verify: running ls (or dir) you see the new empty folder — and Claude Code has opened inside it without errors.

Notice that mkdir e cd need the exact folder name — if you enter a different name in the two commands, the cd it will fail. It’s okay to forget or make a mistake: the terminal displays a message when a command doesn’t work, and you can simply type it again.

Two errors show up at the start every time—and both have simple solutions.

Besides these two, a few specific stumbling blocks come up on the first day for anyone who’s never used a terminal—not one of them is serious, but they can stop you if you don’t know what’s going on.