iOS Performance Agent Skills

A collection of AI-agent skills for reviewing, diagnosing, and improving performance in iOS applications.

Each skill lives in its own top-level directory and installs with a single npx command (see Installation), or can be copied manually into another agent setup.

What this is

This repository contains practical performance-focused instructions for AI coding agents.

The skills are designed to help an agent reason about iOS performance problems with a more structured mental model: what to inspect, what to avoid assuming, which trade-offs matter, and when measurement is required.

It is a collection of reusable guidance files for AI-assisted code review, refactoring, and investigation.

Included skills

ios-launch-performance

For analyzing app startup and launch-time regressions.

Covers cold, warm, and prewarmed launch, dyld and pre-main work, static initializers, AppDelegate, SceneDelegate, SwiftUI app entry points, first frame, SDK initialization, launch metrics, and validation with tools such as Instruments, XCTest, and MetricKit.

ios-performance-profiling

For choosing and interpreting performance profiling workflows.

Covers Instruments, traces, XCTest metrics, MetricKit payloads, signposts, memory graphs, animation hitches, hangs, CPU work, allocations, disk I/O, networking, and production performance signals.

ios-perceived-performance

For improving how fast and responsive an app feels to users.

Covers loading states, first feedback, progressive rendering, skeletons, optimistic updates, high-stakes actions, perceived latency, and validation through recordings, traces, and user-visible behavior.

swiftui-performance

For reviewing SwiftUI code composition and update behavior.

Covers identity, state ownership, dependency scope, body cost, list performance, row complexity, bindings, closures, layout, drawing, animations, async lifecycle work, and profiling validation.

swift-concurrency-performance

For reviewing performance and correctness trade-offs in Swift Concurrency.

Covers actor isolation, MainActor usage, task lifecycle, task groups, cancellation, AsyncSequence, continuations, reentrancy, executor behavior, and concurrency-related responsiveness issues.

swift-runtime-performance

For reviewing Swift runtime-level performance costs.

Covers allocations, ARC traffic, stack vs heap behavior, dispatch, existentials, generics, opaque types, copy-on-write, SIL optimization, unsafe Swift, modularization, linking, and launch-time trade-offs.

Installation

Install the whole bundle with one npx command:

npx skills add Livsy90/iOS-Performance-Agent-Skills --all

Or install only the skills you want:

npx skills add Livsy90/iOS-Performance-Agent-Skills --skill ios-launch-performance --skill swiftui-performance

The CLI will ask which agents to register the skill with (Claude Code, Codex, Cursor, Gemini, …) and whether to install per-project or globally.

If npx is missing, install Node (brew install node); if brew is missing too, install Homebrew first.

Alternative install methods

Claude Code (installs the whole bundle directly):

/plugin install Livsy90/iOS-Performance-Agent-Skills

Or clone this repository and drop the skill folders wherever your agent expects them.

Repository structure

ios-launch-performance/
  SKILL.md
  references/

ios-performance-profiling/
  SKILL.md
  references/

ios-perceived-performance/
  SKILL.md
  references/

swiftui-performance/
  SKILL.md
  references/

swift-concurrency-performance/
  SKILL.md
  references/

swift-runtime-performance/
  SKILL.md
  references/

.claude-plugin/plugin.json    # Claude Code plugin manifest
gemini-extension.json         # Gemini extension manifest