Your first useful prompt

Claude Code installed, interface launched, you find yourself facing a blinking cursor. What do you type? This lesson teaches you how to write a first prompt that works well, and helps you understand what Claude Code is doing under the hood.

Context matters

When you run claude inside a folder, that folder becomes its working context. It can read files, modify them, run commands. That is what sets it apart from a regular chat.

A concrete first prompt

Rather than "hey what can you do", start with a real task. For example:

List the files at the root of the project and tell me what it does.

Type that (or paste it), then submit. Claude Code will call its Glob or Bash tool to list the files, then read the most informative ones (README.md, package.json) before responding.

Understanding what is happening

You will see blocks appear while Claude works:

• Tool calls (Read, Bash, Grep...): these are the actions it is taking.
• Permission prompts: the first time it wants to run a sensitive command, it asks for your authorization.
• Text: what it tells you, outside of any action.

The right reflex

For a prompt to be useful, give it:

1. Context (where you are, which file, in which language).
2. A precise objective (not "improve this", but "rename this function and move it to utils.ts").
3. Constraints (don't touch X, keep Y as-is).