agentsclimarketplace

Native ios orchestrator

Skill Sheshiyer/skill-clusters/skills/native-ios-orchestrator

Hub-and-spoke agent-skill clusters, one per stack (Astro·GSAP·Remotion, Tauri, …). Installable via skills.sh.

Install
npx -y skills add Sheshiyer/skill-clusters --skill native-ios-orchestrator

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Route a native Apple-platform task to the right skill among the iOS specialists — Swift 6.2 concurrency, SwiftUI architecture, actor persistence, protocol-based DI/testing, on-device FoundationModels, Liquid Glass design, and icon assets. USE WHEN a user is building, modernizing, or shipping a native iOS/macOS app in Swift but hasn't named the specific concern.

SKILL.md

9.3 KB, ~2.2k tokens by cl100k_base, as published. Nobody here has run it

Native iOS Orchestrator

The single entry skill for native Apple-platform (Swift / SwiftUI) work. It locates the task on the layer × concern map and delegates to one of 7 specialist spokes. The cross-cutting fact every spoke shares — the iOS 26 / Swift 6.2 / Xcode 26 baseline, its availability-gating discipline, and the data-race-safety model — lives in native-ios-core; read it before adopting the new concurrency mode or any of the iOS 26 frameworks.

Cluster map (spoke → role)

  • swift-concurrency-6-2 — the language model: Swift 6.2 Approachable Concurrency, single-threaded by default, @concurrent for explicit offloading, isolated conformances.
  • swiftui-patterns — the UI architecture: @Observable state, view composition, NavigationStack, list/render performance.
  • swift-actor-persistence — the storage layer: actor-backed in-memory cache + file persistence, data races eliminated by the compiler.
  • swift-protocol-di-testing — the testability seam: small focused protocols for file system / network / external APIs, mocked under Swift Testing.
  • foundation-models-on-device — the on-device AI: Apple FoundationModels LLM, @Generable guided generation, tool calling, snapshot streaming.
  • liquid-glass-design — the design system: iOS 26 Liquid Glass material for SwiftUI, UIKit, WidgetKit.
  • ios-icon-gen — the asset pipeline: PNG icon imagesets for Xcode asset catalogs from SF Symbols or Iconify.

See also ## Folded spokes below for the watchOS and HIG-design spokes.

Routing rules by intent

Set the foundation

  • "Data races / async errors / migrate to Swift 6.2 / MainActor architecture" → swift-concurrency-6-2 (baseline in native-ios-core)
  • "How do I structure views / state / navigation?" → swiftui-patterns

Build the app body

  • Local/offline storage, caching, thread-safe shared state → swift-actor-persistence
  • Make it testable, mock I/O, deterministic tests → swift-protocol-di-testing
  • On-device text generation, structured extraction, AI tool calls → foundation-models-on-device

Make it look native

  • Glass buttons/cards/toolbars, morphing, iOS 26 material → liquid-glass-design
  • HIG-conformant UI, SF Symbols, Dynamic Type, adaptive iPhone/iPad layout, Dark Mode, accessibility → mobile-ios-design
  • App/feature icons, asset-catalog imagesets → ios-icon-gen

Ship to the wrist

  • watchOS app/extension, Watch Connectivity sync, complications, watch workouts/HealthKit, Smart Stack widgets → watchos

Fix / audit / refactor / ship (picked-up workflow spokes)

  • A specific Swift 6.2 data-race / Sendable / actor-isolation compiler diagnostic to fix → swift-concurrency-expert
  • Concrete navigation / sheet / async-.task / deep-link / reusable-screen pattern → swiftui-ui-patterns
  • An existing SwiftUI view is too large / mixes logic with layout / needs splitting → swiftui-view-refactor
  • Janky scrolling, dropped frames, high CPU, hangs, excessive view updates → swiftui-performance-audit
  • Review/implement Liquid Glass against a checklist (availability, modifier order, containers) → swiftui-liquid-glass
  • Run/inspect/debug the app on a booted simulator, capture logs, drive UI → ios-debugger-agent
  • Generate App Store "What's New" / release notes from git history → app-store-changelog
  • macOS menu-bar (LSUIElement) app built with Tuist → macos-menubar-tuist-app
  • Build/sign/notarize/package a SwiftPM macOS app into a .app (no Xcode project) → macos-spm-app-packaging

Folded spokes

These spokes were folded into this cluster from overlapping standalone skills. They share the native-ios baseline (native-ios-core) and the SwiftUI architecture in swiftui-patterns; route to them like any other spoke above.

  • mobile-ios-design — the HIG layer: Apple Human Interface Guidelines made concrete in SwiftUI — layout/grids, navigation patterns, SF Symbols, Dynamic Type, semantic color/materials, and iPhone/iPad adaptivity. Pairs with swiftui-patterns (architecture) and liquid-glass-design (iOS 26 material).
  • watchos — the wrist platform: watchOS apps and Watch extensions — Watch Connectivity (iPhone <-> Watch sync), complications (ClockKit and WidgetKit), HealthKit workout sessions, and Smart Stack widgets. Reuses the cluster's concurrency and SwiftUI conventions on the watch form factor.

Picked-up spokes

These spokes were picked up from the antigravity-awesome-skills library (upstream: Dimillian/Skills, MIT). They are task/workflow spokes — review, fix, refactor, audit, debug, package, ship — that sit on top of the architecture spokes above (swift-concurrency-6-2, swiftui-patterns, liquid-glass-design). Where two spokes overlap, the rule is architecture spoke = how to design it; picked-up spoke = how to fix/audit/ship it. Route to them on demand like any other spoke.

  • swift-concurrency-expert — the concurrency fix workflow: triage a concrete Swift 6.2 data-race / Sendable / actor-isolation diagnostic and apply the smallest behavior-preserving edit, then verify. Pairs with swift-concurrency-6-2 (which is the language model; this one fixes specific compiler errors).
  • swiftui-ui-patterns — the pattern catalog: NavigationStack routing, enum-driven sheets, .task async state, deep links, previews, reusable screens, with a component-reference index. Pairs with swiftui-patterns (architecture) as the concrete how-to layer.
  • swiftui-view-refactor — the refactor workflow: split oversized views into small dedicated subview types, MV-first data flow, stable view trees, side effects out of body. Use after swiftui-patterns/swiftui-ui-patterns when an existing view has grown too large.
  • swiftui-performance-audit — the performance diagnostic: code-first audit of invalidation storms, unstable ForEach identity, layout thrash, and main-thread image decode, escalating to Instruments profiling. Route here for janky scrolling, dropped frames, hangs, or excessive view updates.
  • swiftui-liquid-glass — the Liquid Glass audit/impl: correct glassEffect / GlassEffectContainer / glass button-style usage, modifier order, #available(iOS 26) gating, and morphing. Overlaps liquid-glass-design; prefer this one for review/checklist-driven adoption.
  • ios-debugger-agent — the simulator debug loop: build/run/inspect the current app on a booted simulator via XcodeBuildMCP — UI taps, screenshots, view inspection, runtime log capture. Route here to reproduce a bug or drive a flow at runtime.
  • app-store-changelog — the release-notes generator: turn git history since the last tag into user-facing App Store "What's New" bullets (New/Improved/Fixed), dropping internal-only commits. The ship-time spoke.
  • macos-menubar-tuist-app — the macOS menubar (Tuist) workflow: LSUIElement menu-bar utilities with a Tuist-first manifest/run-script flow and strict model/client/store/view boundaries. The macOS sibling of the iOS architecture spokes.
  • macos-spm-app-packaging — the macOS packaging/release pipeline: scaffold, build, codesign, notarize, staple, and package a SwiftPM macOS app into a .app (plus Sparkle appcast) with no Xcode project. The macOS ship-time spoke.

Standard flow

  1. Locate the task: which layer (language → architecture → storage/test/AI → design → assets) and which concern.
  2. If it touches concurrency isolation, Sendable, availability gating, or any iOS 26-only API, pull the baseline from native-ios-core first — these interlock across spokes (a @MainActor decision in concurrency changes how SwiftUI, actors, and DI are written).
  3. Delegate to the spoke(s). Multi-step asks fan out in layer order (e.g. "build an AI feature with a glass UI" → swift-concurrency-6-2 isolation → foundation-models-on-deviceliquid-glass-design).
  4. Return: chosen spoke(s), the deployment-target / availability implications, and the next action.

Guardrails

See native-ios-core. In short: target Swift 6.2 + Xcode 26, and treat every iOS 26-only API (Liquid Glass, FoundationModels, the newest SwiftUI affordances) as availability-gated — wrap it in if #available / SystemLanguageModel.availability with a real fallback rather than raising the whole app's deployment target silently. Prefer @Observable over ObservableObject, actors over manual locks, and isolated-by-default concurrency over ad-hoc background dispatch. Don't widen a deployment target or weaken data-race safety without saying so explicitly.

Loading spokes on demand

To keep CLI startup context lean, this cluster's spokes are not separately registered as skills — only this orchestrator and its *-core are enumerated. When you route to a spoke named above, load it on demand by reading its file:

~/.agents/skill-clusters/skills/<spoke-name>/SKILL.md (or skills/<spoke-name>/SKILL.md inside the skill-clusters repo).

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.