os-use
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
subprocessmodule to execute AppleScript commands on macOS. This is used for window management, application control, and listing system windows. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external sources and possesses significant system-level capabilities.
- Ingestion points: The skill captures screenshots of the desktop and applications using
pyautogui.screenshot()and reads data from external Excel files usingpandas.read_excel(). - Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present when processing screen content or file data.
- Capability inventory: Includes full mouse and keyboard control (clicking, typing, hotkeys) via
pyautogui, and command execution through AppleScript (subprocess). - Sanitization: There is no evidence of sanitization or validation of the text or visual data processed by the agent before it is used to drive automation logic.
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of several standard, well-known Python libraries from the public PyPI registry, including
pyautogui,opencv-python-headless, andPillow.
Audit Metadata