# Swiftui Performance

> Diagnose and improve macOS SwiftUI rendering, invalidation, table/list, scrolling, image, task, startup, memory, and energy performance using measurement.

- Skill: `rcarmo/swiftui-performance` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add rcarmo/swiftui-performance`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rcarmo/swiftui-performance/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- License: MIT
- Author: rcarmo (https://skillmd.com/u/rcarmo)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/rcarmo/swiftui-performance

---


# SwiftUI performance

Read `references/performance-playbook.md`. Use for slow rendering, scrolling hitches, repeated work, memory growth, startup delay, excessive body updates, battery/thermal issues, or performance-focused review.

## Workflow

1. Reproduce with representative Mac hardware, macOS version, window configuration, data set, and release/debug build as appropriate.
2. Record a baseline metric or trace; do not optimise from code appearance alone.
3. Use Instruments (SwiftUI, Time Profiler, Allocations/Leaks, Core Animation, Network, Energy) and signposts where appropriate.
4. Identify the invalidation source, expensive transform, I/O boundary, allocation, layout cycle, image cost, or task storm.
5. Apply the smallest correction while preserving behaviour and accessibility.
6. Re-measure under the same conditions and add a regression benchmark/test where stable.
7. Build and run on macOS 26 and the lowest-spec supported Mac practical for the product.

## Guardrails

- Do not add `EquatableView`, caches, manual memoization, or flattened view trees without evidence.
- Do not move work to a background task unless ownership, cancellation, actor isolation, and UI commit are correct.
- Avoid `AnyView`, unstable IDs, eager stacks for large data, expensive transforms in `body`, and formatter creation when format styles work.
- Scope timers/`TimelineView` and observation to the smallest subtree.
- Cache only with measured benefit and explicit invalidation/memory bounds.
- Performance changes must not remove labels, text scaling, reduced-motion behaviour, keyboard/focus support, or native macOS semantics.

## Output

Report reproduction, baseline, trace/tool, root cause, change, after measurement, variance, tested hardware/macOS/window state, and remaining risks. Separate measured facts from hypotheses.