SwiftUI Agent Skill: Build Better Views with AI

AI SwiftUI coach drops, devs clap back: “Just build a linter!”

TLDR: An open-source SwiftUI Agent Skill promises AI-built views with fewer mistakes by teaching common best practices. The top reaction demands a proper code-scanning linter instead, kicking off a lively AI-vs-tooling debate that matters to anyone shipping iOS apps fast without weird layout bugs.

An open-source “SwiftUI Agent Skill” just hit GitHub, pitched as a cheat sheet that teaches your AI helper how to build cleaner SwiftUI screens from the start. It bundles a meaty SKILL.md and references for pain points like bad scrolling setups, string formatting, image loading, and that infamous rule: don’t use the one-parameter onChange. The demo claims smarter layouts (no more scroll-in-scroll disasters) and fewer performance hiccups, with the AI citing best practices as it plans. Translation for non-devs: it’s a coaching manual for the robot that writes your app’s screens, so it stops making the same messes.

But the comments lit the fuse. One early reply basically yelled: make the robot scan the code’s structure (think a map of the code) and catch mistakes automatically—aka a linter—rather than spend AI brainpower talking itself into good choices. That split the crowd into two noisy camps: the AI prompt-pack believers who love the quick wins and shared wisdom, and the tooling purists who want static checks that don’t burn tokens or waffle. Jokes flew about AGENTS.md becoming a “confessional” of SwiftUI sins, plus memes about “teaching your robot to clean its room.” Either way, the repo is already attracting pulls from folks trading gotchas—and the vibe is half study group, half fight club.

Key Points

  • An open-source SwiftUI Agent Skill is available on GitHub to help AI agents build and refactor SwiftUI views with better practices.
  • The skill includes a SKILL.md and multiple focused references covering image optimization, layout, list performance, modern APIs, performance, scrolling, sheet navigation, state management, text formatting, and view structure.
  • It enforces guidance such as avoiding the single-parameter onChange variant and identifies issues like nested ScrollView usage and redundant state synchronization.
  • The tool is intended to improve both existing codebases and new agent-generated views; agents may not always discover skills automatically yet.
  • The project accepts community contributions and has already merged several pull requests; the article also mentions Runway as an app release management tool.

Hottest takes

“Have it walk the AST looking for the antipatterns - it ends up more like a custom linter” — cadamsdotcom
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.