What problem does it solve? It verifies whether an Expo skill actually works end-to-end: whether its description triggers it on realistic prompts, whether the code it generates compiles and lints, and whether the resulting app renders correctly on real simulators and emulators. ## Core Features & Use Cases - Trigger accuracy testing: Runs realistic prompts through claude -p subprocesses and measures recall of the skill under test. - With-skill vs without-skill comparison: Generates apps in fresh Expo fixtures with and without the skill, then grades both against expectations. - Runtime screenshots: Runs generated apps in Expo Go or a development build on iOS simulator, Android emulator, or web, and captures screenshots for visual grading against a design rubric or a target reference screenshot. - Use Case: After editing the expo-ui skill's description, run an eval to confirm it still triggers on UI prompts and that the screens it generates render without error overlays on both platforms. ## Quick Start Ask the agent to eval the expo-ui skill with runtime screenshots on iOS and Android using the latest Expo SDK.
