Claude Code Plugins

Community-maintained marketplace

Feedback

>

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name ios-knowledge-skill
description Skill providing structured iOS lane knowledge: SwiftUI vs UIKit/TCA/MVVM patterns, data stack choices, and OS 4.1 iOS standards. Uses context7 libraries for global guidance.

iOS Knowledge Skill – Architecture & Patterns

This skill equips agents with high-level iOS knowledge for the OS 4.1 iOS lane without bloating individual agent prompts.

It is intended for:

  • ios-grand-architect
  • ios-architect
  • Core iOS specialists (SwiftUI/UIKit/persistence/networking/testing/perf/security) when they need architectural context.

Context7 Libraries

When you load this skill, you MAY use context7 MCP tools to read from:

  • os2-ios-architecture

    • SwiftUI vs UIKit vs mixed stack guidance,
    • MVVM/TCA/Clean patterns for iOS,
    • Navigation, modularization, and dependency injection strategies.
  • os2-ios-standards

    • Coding standards for the iOS lane,
    • Error handling and logging patterns,
    • Guidelines for concurrency, testing, and app structure.
  • /websites/developer_apple_swiftdata (via Context7)

    • SwiftData models with @Model, @Relationship, cascade deletes
    • Preferred over Core Data for iOS 17.0+ projects
    • In-memory testing with ModelContainer
    • Native, zero-dependency persistence
  • /ra1028/swiftui-atom-properties (via Context7)

    • Atomic state management for complex cross-feature coordination
    • StateAtom, ValueAtom, ObservableObjectAtom patterns
    • Testing with AtomTestContext
    • Compile-safe dependency injection
  • /pointfreeco/swift-navigation (via Context7)

    • State-driven navigation with @CasePathable enums
    • Sheet/modal presentation tied to model state
    • Deep linking and state restoration
    • Two-way bindings for navigation state

You SHOULD:

  • Read only the sections relevant to the current task,
  • Summarize 3–7 key constraints or patterns into your own working context,
  • Avoid loading large example sections unless explicitly needed.

Usage Pattern

  1. When planning an iOS task:

    • Load ios-knowledge-skill and skim the relevant context7 docs.
    • Extract:
      • Recommended UI stack (SwiftUI vs UIKit/TCA/MVVM),
      • Data strategy hints (SwiftData vs Core Data/GRDB),
      • Lane-specific constraints (design DNA/tokens, concurrency rules).
  2. When writing phase_state.planning:

    • Reflect these decisions in:
      • architecture_path,
      • data_strategy,
      • plan_summary.
  3. When delegating to implementers/gates:

    • Ensure the plan and assignment reflect the chosen patterns,
    • Reference this skill implicitly (do not restate everything in agent prompts).
  4. When planning data persistence (iOS 17.0+):

    • Prefer SwiftData with @Model macro for new projects
    • Use @Relationship(.cascade) for parent-child relationships
    • Plan ModelContainer with isStoredInMemoryOnly for testing
    • Consider SwiftData Query for reactive UI updates
  5. When planning complex state management:

    • Consider SwiftUI Atom Properties for cross-feature state coordination
    • Use StateAtom for mutable state (replaces @State for shared state)
    • Use ValueAtom for computed/derived values
    • Use ObservableObjectAtom for complex objects
    • Plan AtomTestContext tests for state logic
  6. When planning navigation architecture:

    • Consider Swift Navigation for state-driven navigation
    • Model destinations as @CasePathable enums
    • Tie sheet/modal/alert presentation to model state
    • Plan for deep linking via URL → destination state mapping
    • Use navigationDestination(item:) for push navigation

This skill is about progressive disclosure: give agents just enough iOS architecture knowledge to make good decisions, while leaving detailed examples and recipes in context7 where they can be fetched on demand.