The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses the subprocess module to wrap and execute system commands for the Android Debug Bridge (adb) and the Android Emulator. These commands are used to manage the device lifecycle, install/uninstall packages, and interact with the screen.
[DYNAMIC_EXECUTION]: The scripts/build_and_test.py script dynamically locates and executes the gradlew wrapper script within the user's project directory. This involves executing a project-local script whose content is not controlled by the skill itself.
[INDIRECT_PROMPT_INJECTION]: The skill exposes the agent to indirect prompt injection by ingesting untrusted data from the Android device environment.
Ingestion points: scripts/screen_mapper.py and scripts/navigator.py extract UI element text and descriptions from the device's UI hierarchy (XML), while scripts/log_monitor.py streams system logs via logcat.
Boundary markers: The skill does not implement delimiters or specific instructions to ensure that text content from the device is treated strictly as data and not as instructions for the agent.
Capability inventory: The skill possesses significant capabilities, including the ability to execute shell commands on the host and the connected device, install software, and modify device settings.
Sanitization: UI data and logs are retrieved and presented to the agent without sanitization or filtering for potential prompt injection payloads.

android-emulator-skill