Chapter 1Setting up your computer: GitHub, Obsidian and ChatGPT

Chapter 1 · Setup

Setting up your computer: GitHub, Obsidian and ChatGPT

4 min · 778 words

Before you type a single word into ChatGPT you need three things installed and one empty folder. It takes fifteen minutes, and after that you never touch any of it again.

The four tools and what each one is for ChatGPTThe desktop app. Runsthe skill, builds,checks.You: type to itGitHubStores every version ofthe code and publishes.You: create the accountand approveObsidianThe project's memory:everything written downand linked.You: almost nothing(edit copy at the end)VercelFree hosting. Publisheswhatever lands on main.You: import the repoonceThe project folder (the repository)
Figure 4. You only install things and create accounts. ChatGPT uses GitHub and Obsidian for you.

A GitHub account

GitHub is the platform where developers store code. It keeps every version of what we build: if something goes wrong, we can roll back. And on this project it'll do one more thing: it's how the site gets published to the internet.

You won't be using GitHub. ChatGPT will use it for you. All you have to do is create the account and give it access.

  1. Go to github.com and create an account.
  2. Download the GitHub Desktop app (there's a Mac and a Windows version) and install it.
  3. Open it and authorize your account when it asks. On the setup screen, leave everything as it is: it picks up your account details by default, and that's what we need.

Obsidian

Obsidian is also free, and also available for Mac and Windows. Download it from obsidian.md and install it. You don't need to open it now; you do need it installed, because the skill is going to write into it, and later on you'll open it yourself to read the project docs and to change the site's copy.

ChatGPT, the desktop app

This matters: not the browser, the app. And always the latest version. Make sure it's up to date before you start.

The app has a Projects section. For it to let you open a project, you need to be in Work mode. If you don't see the option, switch to that mode.

About the model: in the video I use the most capable model on the "high" setting. It gives better results. "Medium" works too, but when you get to the design steps I'd bump it up to high, because you can tell the difference.

The project folder

This is where everything comes together. Open GitHub Desktop and create a new repository:

  1. At the top, File → New Repository.
  2. Give it a name. I called mine New Website. The default folder is Documents/GitHub; leave it there.
  3. Create the repository and publish it. Make sure "Keep this code private" is checked: your code doesn't need to be public.

If GitHub Desktop asks for system permissions when you publish, grant them. It's the first time the app is doing anything and the operating system wants you to sign off on it.

Now open ChatGPT, go to Projects, and open that folder: New Website. It's empty. That's how it should be. Give the project a name so ChatGPT knows what it's about, and when it asks whether you trust the folder, say yes.

Downloading and installing the skill

The skill lives in a public GitHub repository. There are two ways to hand it to ChatGPT and both work.

The simplest is to give it the link. In the project you just opened, type:

Install this skill: https://github.com/themariscalproject/themariscal-create-amazing-website-skill

That's it. ChatGPT will read the repository, think for a bit, and install it.

The other way is to download the repository as a ZIP, unzip it in Downloads, and drag the folder into ChatGPT with the same message: "install this skill." That's what I do in the video.

Once it's done, the skill is available in all your projects. To run it, at any time, type a slash (/) and a list of skills pops up. Search for "mariscal" and there it is: themariscal-create-amazing-website.

The permissions it's going to ask for

On the first real step, when the skill tries to push things to GitHub, two things will happen that look a bit alarming and shouldn't.

First, ChatGPT will say it can't publish to GitHub and needs authorization. Give it, and give it permanently. It's so it can send the code to your repository, which is what lets you keep track of everything that gets done.

Second, on a Mac the system will ask you to enable two things under Privacy & Security: Accessibility and Screen Recording. Drag the app into the list, enter your admin password, done. Why does it want to record the screen? To take screenshots. ChatGPT will screenshot what it's building to check that it looks right, especially on a phone. It's one of the things that makes the result good.

This happens once. After that it won't ask again.

With that, your computer is ready. Next up: telling it the idea.