Android screen observation
Capture and describe the device display: single frames, recordings, and the display configuration that every coordinate-based skill relies on.
When to use
- Seeing what is currently on the device (before/after an action).
- Recording a UI session as evidence.
- Reading resolution/density/orientation before
android-inputorandroid-ui.
Tools used
take_screenshot— capture the current frame (optionally region-cropped).screen_record— record a bounded screen session to a file.screen_info— resolution, density, orientation, rotation, display modes.
Procedure
screen_infofirst when coordinates will be involved.take_screenshotto observe current state; store the frame as evidence.- For multi-step observation, capture before and after each mutation.
screen_recordonly for sessions the human explicitly requested.- Never trust a stale frame: re-capture before relying on screen state.
Security
- Captured frames can contain private content (messages, tokens, photos): keep them local, task-scoped, and delete on completion.
- Screen recording is mutative to storage and privacy:
confirm=true.
Related skills
android-ui · android-input · android-vision · android-device-info