device-interaction
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the application's UI hierarchy and screenshots, creating a vulnerability where an attacker-controlled application could display content designed to manipulate the subagent's logic.\n
- Ingestion points: UI hierarchy files (containing element names and values), screenshots, and application runtime logs ingested via
DeviceEventSynthesizeandDeviceInteractionInstallAndRun.\n - Boundary markers: The instructions do not define delimiters or provide specific warnings for the agent to ignore instructions found within the retrieved UI elements.\n
- Capability inventory: The skill can perform complex device interactions including synthesized touch events, hardware button presses, and keyboard input via the
DeviceEventSynthesizetool.\n - Sanitization: The instructions do not specify any sanitization or validation of content retrieved from the device's UI before it is processed by the agent.
Audit Metadata