The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates screenshot capture by executing various system utilities and bundled scripts across different operating systems. This involves calling tools such as screencapture on macOS, scrot, gnome-screenshot, and import on Linux, and osascript for application management. These commands are executed via subprocess.run in Python or directly in shell scripts.
[PRIVILEGE_ESCALATION]: The skill's documentation instructs the agent to execute the Windows capture script using powershell -ExecutionPolicy Bypass. This command overrides the system's default script execution policy, which is a form of privilege escalation intended to ensure tool functionality.
[INDIRECT_PROMPT_INJECTION]: The skill captures pixel data from the user's environment, which serves as an ingestion point for untrusted data. Malicious content visible on the screen during a capture (e.g., within a browser or document) could provide instructions that influence the agent's subsequent analysis or actions.
Ingestion points: scripts/take_screenshot.py and scripts/take_screenshot.ps1 capture screen buffer data into image files.
Boundary markers: Absent. The skill produces raw images without explicit markers to distinguish user-controlled content from system or application data.
Capability inventory: The skill possesses the capability to write files to arbitrary user-specified or temporary paths and execute system-level commands for screen capture and application focus.
Sanitization: The skill performs basic sanitization by escaping double quotes in application names used in AppleScript and validating pixel coordinate inputs for regions.

screenshot