Claude Code Basics

Before you panic

When you first start, things are going to look a little weird. Unfamiliar words, popups you've never seen before. That's completely normal, this page translates all of it.

The most important thing: Claude always asks before it does anything. You're in control the whole time. Not sure about something? Just say no. Claude will figure it out.

Permission popups

Claude Code will ask your permission before it takes certain actions. Think of it like a new employee checking with you before doing something. Here's what each one means.

Read (always safe). "Do you mind if I take a look at this document?" Claude is just reading, not changing anything.

Write (review first). "Do you mind if I create a new document for you?" Take a quick look at the file name to make sure it looks right, then say yes or no.

Edit (review first). "Do you mind if I edit this document? I'll show you exactly what I'm changing before I do it." You see the before and after, so you know exactly what changed.

Bash (review first). "Do you mind if I run this program for you?" Don't let the word "Bash" scare you. It's just the name of the tool that runs commands on your computer. This is how Claude installs things, runs programs, and actually gets work done.

The golden rule. When in doubt, click No. You can always try again. Claude will explain what it was trying to do and suggest a different approach. Nothing bad happens from saying no.

When Claude runs commands

Sometimes Claude asks to run something that looks like total gibberish. Here's what it's actually asking to do.

Installing things

• npm install (safe). Downloads the tools this project needs to run. Like grabbing an app from the App Store, comes from a trusted public library.
• pip install (safe). Same idea, just for Python projects. You don't need to know Python. You're just giving the thumbs up to download.

Looking at files

• ls (always safe). Just peeking at what's in a folder.
• cd (always safe). Opening a different folder. That's it.
• cat (always safe). Reading a file. Not touching it.

Saving your work

• git add (safe). Flagging which files to save. Like setting them in a pile before hitting Save.
• git commit (safe). Taking a snapshot of where you are with a little note. Your personal undo history.
• git push (review first). Uploading your saved work to the cloud. Worth a quick check before saying yes, once it's up, it's up.

Building and running

• npm run build (safe). Packaging everything up so it's ready to go live.
• npm run deploy (review first). This is the one that puts it live. Take one last look before saying yes.
• python (safe). Claude is running a small program to handle something for you, sending emails, sorting files, processing data. You don't need to know Python.

Words you'll see (and what they mean)

These terms come up a lot in Claude Code. Here's what they mean in plain English.

• Terminal. The dark screen with all the text. It looks intense at first, but it's just a window you can type in.
• Bash / Shell / PowerShell. Different names for the same basic idea: a program that lets you type instructions to your computer.
• Git. Your project's entire history, saved as you go. If something goes sideways, you can roll back. It's "undo" for your whole project.
• Repository (repo). A project folder that Git is watching. When someone says "repo," they just mean the folder.
• npm. The thing that downloads tools for web projects. Stands for Node Package Manager, but you never need to think about that again.
• pip. Same deal as npm, just for Python projects.
• API key. Basically a password that lets one app talk to another. Treat it like a real password.
• .env file. A hidden file where your passwords and API keys live. The dot at the front keeps it hidden. Claude knows not to touch it.
• Dependencies. The add-on tools your project needs. Like setting up a new phone with all your apps.
• Deploy. Putting your project live so anyone can see it.
• Script. A file with a list of steps for the computer to follow automatically. A recipe, basically.
• Glob. A shorthand way to describe which files Claude should look at. Instead of listing every file, you write a pattern like "everything in this folder that ends in .js."

Common file types

• .html, a web page. What your browser reads to show you a website.
• .css, styling for a web page. Colors, fonts, layout.
• .js, JavaScript. Makes web pages interactive.
• .py, a Python file. Python is great for automating tasks.
• .json, a data file. Like a spreadsheet for computers.
• .md, a Markdown file. A simple text document with basic formatting.
• .env, stores private settings like passwords and API keys. Never share this file.

Skills vs. Agents

You'll hear these two words a lot. They sound similar but do different things.

What's a skill? A skill is a saved set of instructions Claude can follow. Think of it like a recipe card. You create it once, then any time you need that thing done, you say "use this recipe" and Claude follows the steps. A skill is a shortcut. It saves you from repeating yourself.

What's an agent? An agent is Claude working on a bigger task on its own. Instead of telling it each step, you give it a goal and it figures out how to get there. It might use multiple skills, read files, run commands, and make decisions along the way.

The easy way to remember. Skill = a recipe (specific instructions for one task). Agent = a chef (uses recipes and judgment to complete a bigger goal).

Your safety net

Here's what's actually protecting you.

Claude always asks before it does anything. Change a file, run a command, install something, you see it before Claude touches it.

You can always say no. It won't break anything. Claude will just find another way, or ask what you'd prefer.

Git is your undo button. Every change Claude makes gets saved in the history. If something goes sideways, you can go back. Nothing is permanent.

Your passwords stay private. Claude knows where your .env files and API keys are, and won't touch them or share them.

When to pause and think. If you see the words delete, remove, or force in something Claude is asking to run, take a second to read it. Claude will usually explain why, but those are worth an extra look.

Quick cheat sheet

Safe to approve. Read, ls, cd, cat, git add, git commit, npm install, pip install, npm run build, python.

Glance before approving. Write, Edit, git push, npm run deploy, anything that says "send" or "publish."

Read carefully first. Anything with "delete", "remove", "force", or "reset". These are rare and usually fine, but read the full command first.

Still have questions? Join The Owner's AI Room (free community) and ask anything. Or book a call if you want personalized help.