android-adb
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill makes extensive use of shell command templates where user-provided or environment-provided variables (e.g., SERIAL, package, activity, coordinates, or the raw command field) are interpolated. If the agent populates these fields with unsanitized data from untrusted sources, it could lead to command injection on either the host machine or the connected Android device.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Android device via commands like
adb shell dumpsys windowandadb shell uiautomator dump. - Ingestion points: Device metadata and UI hierarchy dumps are retrieved in
SKILL.md. - Boundary markers: No specific delimiters or "ignore" instructions are provided for the output of these commands.
- Capability inventory: The skill can execute shell commands (
adb shell), install packages (adb install), and pull files (adb pull). - Sanitization: No evidence of sanitization or validation of the device-provided strings before they are processed by the agent.- [DATA_EXFILTRATION]: The skill includes commands to capture screenshots (
screencap) and dump UI hierarchy data (uiautomator dump), which are then pulled to the local machine. This constitutes access to potentially sensitive information stored on the device or displayed on the screen.
Audit Metadata