android-cli
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches installation scripts and updates from a trusted distribution domain (dl.google.com).
- [REMOTE_CODE_EXECUTION]: Provides procedures to download and execute shell scripts for setting up the Android CLI environment. The skill recommends a manual review of the script content before execution to mitigate risks from mutable network content. Additionally, the
updatecommand accepts a user-provided URL for downloading updates. - [COMMAND_EXECUTION]: Executes shell commands via
adband theandroidCLI to manage SDKs, create projects, and interact with emulated or physical devices. - [PROMPT_INJECTION]: Exposure to indirect prompt injection through the processing of untrusted data from application UI layouts and XML 'journey' test files. The ingestion points include layout JSON and XML action text. The instructions include boundary checks, requiring the agent to verify that journey actions are valid UI interactions before execution.
- [DATA_EXFILTRATION]: Accesses sensitive information from connected devices, including screen contents and UI hierarchy data, which are used to inform agent actions and test results.
Audit Metadata