Full course · step 1 of 4 · 8 lessons
From chat to the terminal: the agent working in your folder
Open the terminal without fear, install the agent from the official source, set up the folder with rules, save each version with Git, keep keys out of the text, and put your page online. Command by command, with what to check at every step.
- 8lessons
- 42templates with a copy button
- 6practice terminals
- 17exercises with a done-when checklist
The lessons
In the order you’ll do them. Each lesson ends with practice and a short test; move on only when the “done when” condition is met.
Lesson 1 · 15 minThe terminal starts by telling you where you areThe terminal is just another window into the same folders you see in File Explorer. You type one line, and it returns another. With six commands, you can see where you are, what’s there, create your projects folder, and move between folders without a mouse.
Lesson 2 · 18 minInstall from the official source and sign in with your accountInstalling the program, signing in to your account, and opening it in the folder are three separate steps, each with its own check. The command comes from the official page, the account is your subscription, and no password passes through the terminal.
Lesson 3 · 16 minThe folder is the context: README, AGENTS.md, and MarkdownWhen everything is in one folder, the AI reads material that has nothing to do with your request, and you can’t tell where each sentence came from. One folder per topic, with one file for people and another for the agent, solves this before your first request.
Lesson 4 · 18 minFirst the agent reads, then it makes changes“Fix the project” mixes diagnosis and changes: if something goes wrong, you won’t know which part caused it. Separating reading from changes, giving the minimum permission needed, and asking for a wrap-up that says what was checked turns the agent into someone you can manage.
Lesson 5 · 14 minKeep secrets out of the textA key pasted into a document, a screenshot, or a request gives access to anyone who reads it, and it stays valid after the conversation ends. Separating the secret from everything else before saving the first version lets you share the folder without worry.
Lesson 6 · 20 minGit: the safety net for undoing any changeFor a task, an agent can change ten files at once. Without history, you don’t know what changed or how to go back. With Git, each good version is saved with a date and explanation, and you can undo a change you didn’t approve with one command.
Lesson 7 · 18 minAGENTS.md, Skills, and memory you can checkEvery new conversation with the agent starts from scratch. Without an instruction file, you repeat the same reminders; with vague instructions, nothing changes. Rules you can check, a procedure saved in a Skill, and short, dated memory help the agent improve with every task.
Lesson 8 · 20 minPage live and the same project on a VPSSaving, sending, and putting it online are three steps and three decisions. With an SSH key and gh, you send the project to GitHub, publish the page at an address that opens on your phone, and bring the same folder to a VPS, where the agent works even when your computer is off.
Who it’s for
- 1People who use ChatGPT, Claude, or Gemini in a browser and have never opened a terminal.
- 2People who have seen the program overview and want to complete step 1 for real, with every command.
- 3People who will guide a team or client and need to know how to check what the agent did.
What you’ll be able to do
- 1Navigate folders with cmd (and the Mac or Linux Terminal), always knowing where you are.
- 2Install and sign in to Claude Code or Codex with your subscription, checking the sign-in method.
- 3Set up the agent’s folder with README, AGENTS.md, .env outside Git, and history you can roll back.
- 4Publish the page on GitHub Pages and move the same project to a VPS.