vmos-edge-control-api

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands (via curl) to interact with the Android Control API. It also provides an explicit interface for executing shell commands on the target Android device through the /system/shell endpoint, including optional root access.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Android device UI to drive automation logic, creating a potential surface for indirect prompt injection if the device displays attacker-controlled content.
  • Ingestion points: Data is ingested via /accessibility/dump_compact, /activity/top_activity, and screen analysis in references/observation.md.
  • Capability inventory: The agent can execute shell commands, install applications via URI, and inject text/input based on observed UI content.
  • Boundary markers: No specific delimiters or instructions to treat device data as untrusted are present in the prompt instructions.
  • Sanitization: The skill does not specify sanitization or validation logic for the content retrieved from the device's accessibility services.
  • [EXTERNAL_DOWNLOADS]: The skill documentation and installation instructions reference resources and code from the vendor's official GitHub organization (vmos-dev/vmos-edge-skills).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 02:41 AM
Security Audit — agent-trust-hub — vmos-edge-control-api