Apple Design Skill turns an agent into a UI reviewer that reads Apple’s Human Interface Guidelines before it judges a screen. It carries 123 guideline pages from developer.apple.com plus a curated Liquid Glass guide, and translates Apple’s vocabulary for SwiftUI, UIKit, AppKit, Flutter, React Native, Tauri and Electron.
A review runs through five lenses in order: accessibility, platform conventions, visual craft, interaction and writing. Each finding gives what is wrong, the guideline file and heading that supports it, and a fix in your framework. A craft lens also asks whether the design has a point of view or looks like a template.
What You Get
- A structured design review with Summary, Critical, Improvements, Craft notes, What works and Platform notes sections
- Citations to the guideline file and heading, such as buttons.md under Style, for every finding
- Measured numbers for contrast, type size and control size in each finding
- Specialized modes for app icons, dark mode, Liquid Glass, Apple Pay, navigation, iPhone Duo, onboarding, forms, generative AI and widgets
- An improvement mode that plans a compact token system and a signature element before proposing fixes
Common Use Cases
- Review a Flutter or React Native screen against iOS conventions
- Audit a Tauri or Electron app for macOS conventions and accessibility
- Check a repository for Apple Pay buttons that App Review would reject
- Make a generic-looking interface more distinctive without breaking platform patterns
- Check a layout for the two-display iPhone Duo
What Makes It Unique
The guidelines are stored in the skill as Markdown pages in Apple’s own wording, so the agent quotes them from the file it opened. A routing table names which pages to load for what is on screen, and a heritage file traces each current principle to Apple’s archived guidance from 1980 to 2014 and lists advice that has since been replaced. The craft lens names three looks that dominate generated interfaces and asks whether a design chose its palette and type for a reason.
Under The Hood
The agent works from the files in the skill: SKILL.md, the references folder with the guideline pages, a cross-platform translation file, a code-audit file and a heritage file. For a code audit it searches your source with grep-style commands and reads each match. The repository also holds a Node script that re-downloads the guidelines from Apple’s site, about 170 requests, and rewrites the reference files; it is for maintainers and a review does not need it. Without that script the skill reads files and runs nothing on its own.