mobilerun-cloud
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses curl to interact with the Mobilerun API at api.mobilerun.ai for device management, screen observation, and automation actions.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages sensitive data including device screenshots, accessibility trees, and stored app credentials. It includes strict rules instructing the agent to never reveal the MOBILERUN_API_KEY in chat and to protect the privacy of screen content.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from mobile device screens, creating a surface for indirect prompt injection. 1. Ingestion points: The UI state (accessibility tree) and screenshot endpoints retrieve external content from the device. 2. Boundary markers: The instructions recommend using targeted jq filters to isolate data, but do not specify mandatory prompt delimiters to prevent the agent from following instructions embedded in the UI. 3. Capability inventory: The agent has capabilities for UI input simulation, file transfer, and shell command execution on the target mobile device. 4. Sanitization: Screen-extracted content is processed as retrieved without explicit sanitization steps for input validation.
- [REMOTE_CODE_EXECUTION]: The documentation references official resources for setup, including the Portal APK from the droidrun GitHub repository and the droidrun Python package on PyPI. These are standard components of the vendor's ecosystem and are not used for unauthorized remote code execution.
Audit Metadata