argent-vega

Pass

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to use shell commands (vega, npm, vsm) for managing virtual devices, configuring developer mode, and handling port forwarding. These are standard operations within the Fire TV (Vega) development ecosystem.
  • [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it processes untrusted data from external applications via UI element trees.
  • Ingestion points: The describe tool, which reads the on-screen element tree of any active application.
  • Boundary markers: Absent; the skill does not define specific delimiters for separating UI content from instructions.
  • Capability inventory: The skill possesses extensive capabilities including app reinstallation, input simulation (D-pad and keyboard), and host-side shell execution.
  • Sanitization: The instructions do not specify any sanitization or validation logic for the content retrieved from the UI tree.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 26, 2026, 08:22 PM
Security Audit — agent-trust-hub — argent-vega