What problem does it solve? AI-generated macOS SwiftUI code often misuses the scene graph at App.body level: Preferences crammed into the main window instead of a Settings scene, menu-bar apps faked with NSStatusItem, openWindow calls that silently no-op, and the macOS 26 MenuBarExtra activation trap where windows open invisibly. This Skill systematically detects those defects in an existing project. ## Core Features & Use Cases - 13-rule defect index (sw-01 to sw-13): Covers stale or invented scene APIs, missing Settings scenes, modal Save/Cancel settings windows, sheet-faked auxiliary windows, NSStatusItem misuse, the MenuBarExtra activation trap, unmatched openWindow ids, missing dismissWindow paths, absent lifecycle bridges, and wrong window chrome. - Hybrid lint engine: Runs tier-1 ripgrep tells plus tier-2 ast-grep structural rules (containment checks grep cannot express), emitting unified JSON and SARIF with a per-file parse probe and CI exit codes. - Evidence-grounded fixes: Every finding is verified against a corpus of 1,857 real macOS apps via the swiftui-ctx CLI and Apple docs fetched through Sosumi, with permalinked real-code examples backing each recommended fix. - Use Case: Point it at a macOS SwiftUI project whose menu-bar app's Settings button does nothing on macOS 26; it locates the openSettings call inside the MenuBarExtra closure, flags the activation trap, and reports the hidden-Window workaround with citations. ## Quick Start Audit the SwiftUI sources in this macOS project for scene and window defects and write findings under swiftui-audits/scenes-windows.
