device
the device whisperer
device work is evidence work: drive the app, observe the real screen, and collect what you saw. the platform (ios simulator, android emulator, tv, cloud) changes the commands, not the discipline. automation core first, per-platform recipes on top of it, then exploratory testing and maestro cloud runs.
real screens, real tapsreproduces before fixingrecords the run
Driving, verifying, and testing apps on simulators, emulators, or real devices: navigation, taps, screenshots, UI state extraction, Maestro flows, dogfooding, and failure evidence.
methodology
- identify the platform and whether the app is already installed and running.
- reach for agent-device first: it is the shared automation core for navigation, input, snapshots, and evidence; ios-simulator and android-emulator are the per-platform recipes over it.
- “dogfood this app” routes to dogfood for the exploratory QA method.
- for unfamiliar Maestro syntax, consult the installed
maestro --helpor official documentation; cloud runs go through limrun-maestro-testing. - live project
AGENTS.mdand operator directives override everything here.
contents
- trainedlearned layer: read first
- agent-deviceautomating Apple-platform, Android, and Vega OS TV apps: navigation, taps, typing, scrolling, snapshots, UI extraction, evidence
- android-emulatorverify and debug native/RN/Expo/Flutter apps on an Android virtual device with agent-device
- dogfoodsystematic exploratory testing of a mobile app to find bugs, UX issues, and problems
- ios-simulatorverify and debug native/RN/Expo/Flutter apps on iPhone/iPad Simulator with agent-device
- limrun-maestro-testingrun/write/debug Maestro YAML flows on a Limrun cloud iOS simulator from any environment