software-android-design
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides utility scripts located in the
scripts/directory that wrap standard Android developer commands.scripts/run-android.shandscripts/build-android.shexecute./gradlewto build and deploy applications, whilescripts/capture-screenshot.shandscripts/layout-inspector.shutilizeadbto extract UI data and screenshots from connected devices. - [EXTERNAL_DOWNLOADS]: The
scripts/bootstrap-emulator.shscript automates the installation of Android system images usingsdkmanager. This process fetches resources from official Google repositories and is a standard requirement for creating virtual testing environments. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from the UI of audited applications. Ingestion points: UI hierarchy XML and screen captures retrieved from the emulator via ADB. Boundary markers: Not implemented; scripts pass raw device output to the filesystem. Capability inventory: The agent has the ability to write files and execute local shell commands through the provided script suite. Sanitization: No filtering or sanitization of UI text content is performed, which is a common surface for indirect injection in audit-focused agents.
Audit Metadata