An Instruments-wielding SwiftUI expert for your AI agent

What it does

This repository packages SwiftUI expertise into the Agent Skills open format so AI coding assistants can reference best practices on demand. It covers state management, view composition, navigation, animations, macOS scenes, and iOS 26+ Liquid Glass through modular reference files that load only the relevant topic, keeping the agent’s context window uncluttered.

The interesting bit

The standout addition is a pure-Python toolchain that wraps Apple’s xctrace to record and analyze Instruments traces. The parser correlates hangs and animation hitches with main-thread samples across five diagnostic lanes, then emits structured JSON and markdown so the agent can diagnose SwiftUI performance issues without drowning in raw xctrace export output. A single metric—main_running_coverage_pct—quickly separates CPU-bound work from blocked-main-thread problems.

Key highlights

• Covers the full SwiftUI surface: lists, sheets, Swift Charts, accessibility, image optimization, and deprecated-to-modern API migration
• Reference files load on demand, so agents get deep guidance without context bloat
• Bundled Python scripts for recording and analyzing .trace files, with time-window scoping and signpost discovery
• Includes a maintenance skill to refresh API guidance after new iOS or Xcode releases
• Non-opinionated focus on correctness and performance rather than architecture prescriptions

Caveats

• The trace-analysis toolchain requires xctrace, which ships with Xcode
• The maintenance skill for updating deprecated APIs depends on the external Sosumi MCP tool

Verdict

SwiftUI developers who want their AI assistant to reference concrete best practices instead of improvising should look here. Skip it if you’re not using an agent that supports the open skill format.