The Agent Skills Directory

[COMMAND_EXECUTION]: The Python and shell scripts execute various system utilities like screencapture, scrot, gnome-screenshot, import, osascript, and swift using subprocess.run. This is required for the skill's primary function of capturing screenshots across different operating systems.
[DYNAMIC_EXECUTION]: The tool executes Swift source code files (.swift) at runtime using the swift interpreter to interact with macOS APIs for window and display information. Additionally, the PowerShell script uses Add-Type to compile and run C# code for accessing Windows native methods.
[PRIVILEGE_ESCALATION]: The documentation mentions using -ExecutionPolicy Bypass for the PowerShell script and suggests rerunning with higher permissions if the sandbox blocks access. While this involves bypassing local execution policies, it is a standard requirement for running unsigned administrative or utility scripts in a Windows environment.
[INDIRECT_PROMPT_INJECTION]: The skill provides functionality to list and select windows by name, creating a surface for indirect prompt injection.
Ingestion points: scripts/take_screenshot.py reads application and window names from the operating system via scripts/macos_window_info.swift.
Boundary markers: The skill output prints paths and window information without specific delimiters or instructions for the agent to ignore embedded commands.
Capability inventory: The skill has extensive capabilities including shell command execution via subprocess.run (Python), bash (Shell), and C# compilation (PowerShell), as well as file system write access.
Sanitization: Application names are partially sanitized by escaping double quotes before being used in AppleScript, but window titles are not sanitized before being returned to the agent context.

screenshot