0 min read

Generate a Type Scale From Scratch

Designers

Claude Skill

TL;DR:

Setting up type for a new design system is tedious. Pick a scale, size every heading and body level, wire it all into variables, do it again per breakpoint. None of that needs taste, it needs patience and consistency. I built this as a Claude skill that runs the setup, previews the result, then builds the variables and text styles in Figma.

Why This, Specifically

"AI replacing designers" is the wrong framing for most of the job. Taste and judgment aren't what this touches. What it replaces is the part that was never about taste, computing values off a ratio, naming them consistently, wiring them into styles, repeating it per breakpoint. That's the setup labor people actually mean when they talk about automating repetitive work. It builds the foundation, you take it from there.

How It Works

You give it a base size, a typeface for heading and body, and a scale ratio. It computes every size, snaps them to a shared token scale so type isn't its own island separate from the rest of your system, and shrinks headings by a fixed rule across breakpoints. Everything lands in Figma-native units, since Figma doesn't actually support rem or pt on these properties.

The Skill

It runs the full sequence: setup questions, a real preview you approve before anything touches Figma, then the variables, the text styles, and a specimen frame so you can see it applied, not just described.

When and How To Use It

New type scale, new project, new client, nothing existing yet. Ask for something like "set up a type scale for a new project" and it walks you through it. It's not for updating an existing scale, it'll tell you that and stop.

A Note on What This Skill Can and Cannot Do

Initial creation only. It can get things wrong. Check it twice: once at the preview, before anything reaches Figma, and again after the build, in the actual Variables panel and styles, not just the summary it gives you when it finishes.

This skill is a structured prompt. It instructs Claude to evaluate without softening criticism, but it does not guarantee accuracy. Claude can still miss things, misread context, or produce confident-sounding findings that turn out to be wrong. The output is a starting point for your own judgment, not a substitute for it.
Treat any finding that would drive a real decision as something to verify independently before acting on it.

Get the Skill

The skill is available free on GitHub.

Alex Dihel | Product & Marketing Design Leadership | Design Operations   www.alexdihel.com © | Privacy

Alex Dihel | Product & Marketing Design Leadership | Design Operations   www.alexdihel.com © | Privacy

Alex Dihel | Product & Marketing Design Leadership | Design Operations   www.alexdihel.com © | Privacy