running-mobile-simulators

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill manages mobile virtual devices by executing shell commands such as adb, emulator, xcrun, and maestro. These commands interpolate variables like <avd-name>, <serial>, and <udid> directly into shell strings.
  • [INDIRECT_PROMPT_INJECTION]: The skill exposes a command injection surface through the processing of device identifiers and automation flow paths which may be sourced from the local workspace.
  • Ingestion points: AVD names, UDIDs, and Maestro flow paths defined in workspace configuration or requested by user input (SKILL.md, references/android.md, references/ios.md, references/maestro.md).
  • Boundary markers: Absent. The skill lacks instructions to sanitize or validate these identifiers before shell interpolation.
  • Capability inventory: Subprocess execution of mobile development CLI tools (adb, xcrun, maestro) across all reference scripts.
  • Sanitization: Not provided. The skill does not instruct the agent to escape shell metacharacters in identifiers.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 04:28 PM
Security Audit — agent-trust-hub — running-mobile-simulators