device-interaction

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes UI hierarchy data and screenshots from external applications, creating a potential vector for indirect prompt injection if the application UI contains malicious instructions.
  • Ingestion points: DeviceEventSynthesize captures UI hierarchy text and screenshots in SKILL.md.
  • Boundary markers: No explicit markers or instructions are provided to the subagent to distinguish between UI text data and control instructions.
  • Capability inventory: The skill allows synthesizing touch events, hardware button presses, and keyboard input (DeviceEventSynthesize), as well as launching apps with custom environment variables and arguments (DeviceInteractionInstallAndRun).
  • Sanitization: There is no evidence of sanitization or filtering of the UI hierarchy content before it is read by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 04:12 PM
Security Audit — agent-trust-hub — device-interaction