screenshot

Pass

Audited by Gen Agent Trust Hub on May 11, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes various system utilities to perform its primary function. On macOS, it calls screencapture, osascript, and defaults. On Linux, it invokes scrot, gnome-screenshot, or import. These are standard tools for screen capture on their respective platforms. Additionally, the Python script executes bundled Swift scripts via the swift interpreter, and the PowerShell script uses Add-Type to compile inline C# for interaction with the Win32 API.
  • [DATA_EXFILTRATION]: While the skill accesses sensitive screen information, it only saves the output to local files specified by the user or the system's temporary directory. No network operations or external communication patterns were found that would indicate data exfiltration.
  • [EXTERNAL_DOWNLOADS]: All logic is contained within the skill's bundled scripts (Python, PowerShell, Swift, and Shell). It does not download external code or dependencies at runtime.
  • [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection if the agent subsequently processes the content of the screenshots.
  • Ingestion points: Screen captures are saved to local files (scripts/take_screenshot.py, scripts/take_screenshot.ps1) and then reviewed by the agent.
  • Boundary markers: None; the skill captures raw visual data from the display.
  • Capability inventory: The skill can execute system capture binaries and write images to the local filesystem.
  • Sanitization: None; the content is captured as displayed on the user's screen.
Audit Metadata
Risk Level
SAFE
Analyzed
May 11, 2026, 02:49 PM