screenshot

Run the take_screenshot.py helper with the --app flag followed by the application name, which captures all matching on-screen windows. Use --list-windows first to discover window ids, then --window-id for a precise single-window capture.

Pass a region as x,y,w,h pixel coordinates, for example --region 100,200,800,600. On macOS it maps to screencapture -R, on Linux it uses scrot -a or ImageMagick import with crop geometry, and on Windows the PowerShell script accepts -Region.

macOS requires Screen Recording permission before any capture works. Run the ensure_macos_permissions.sh preflight script once to request it, and if you see sandbox or Swift ModuleCache errors, rerun the command with escalated permissions.

The helper automatically selects the first available tool in this order: scrot, gnome-screenshot, then ImageMagick import. Region captures require scrot or ImageMagick, and active-window capture can also use import combined with xdotool.

Yes, Windows capture is handled by the take_screenshot.ps1 PowerShell script rather than the Python helper. It supports default, temp, and explicit output paths plus region, active-window, and window-handle modes using System.Drawing.

Prefer tool-specific capture such as Playwright for browsers or a Figma integration for design files when available. Use OS-level capture for whole-desktop screenshots, desktop apps without integrated tooling, or when tool-specific capture cannot reach the target.