The Agent Skills Directory

[SAFE]: The skill performs its described functionality using native platform tools: screencapture on macOS, scrot, gnome-screenshot, or import (ImageMagick) on Linux, and native GDI+ APIs via PowerShell on Windows. No indicators of malicious intent or data exfiltration were found.
[COMMAND_EXECUTION]: In take_screenshot.py, the script utilizes subprocess.run to call system utilities. It follows security best practices by passing arguments as lists rather than shell strings, which mitigates shell injection risks. The script also includes an activate_app function that uses osascript to bring target applications to the foreground, employing quote-escaping as a basic sanitization measure for application names.
[DYNAMIC_EXECUTION]: On macOS, the skill leverages bundled Swift files (macos_permissions.swift, macos_window_info.swift, and macos_display_info.swift) to interact with system APIs for permission checks and window discovery. These are compiled and executed at runtime via the swift interpreter, which is a standard approach for macOS automation skills requiring access to CoreGraphics and AppKit frameworks.
[DATA_EXPOSURE]: While the skill accesses screen content, it only does so upon explicit request. The resulting image files are saved locally to paths specified by the user, a temporary directory, or the operating system's default screenshot location. No network transmission or external data exfiltration logic is present in any of the scripts.

screenshot