argent-device-interact

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is designed for mobile development and automation tasks. It provides documentation and tool references for common simulator/emulator interactions such as tapping, scrolling, typing, and app management.
  • [DATA_EXFILTRATION]: Although the skill includes tools for capturing screenshots and screen descriptions (describe), these are standard functionalities for UI automation and discovery. They do not involve unauthorized access to or transmission of sensitive data from the host machine.
  • [COMMAND_EXECUTION]: The commands referenced (e.g., adb, simulator service calls) are focused on the controlled environment of a mobile simulator or emulator. There is no evidence of arbitrary or dangerous system command execution.
  • [PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection via device screen ingestion (reading UI element labels or screenshots), which is inherent to the nature of a UI automation tool. However, the instructions encourage best practices and the use of discovery tools that provide structured data.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 06:25 PM
Security Audit — agent-trust-hub — argent-device-interact