screenshot

TL;DR

A screen-capture skill from OpenAI's catalog of Codex skills. It follows three save rules every time: your path if you give one, the OS's default screenshot location if you just ask for a screenshot, and the temp directory when the agent captures something to look at itself. On macOS a preflight script checks Screen Recording permission and asks for it in one place, and the agent chains it with the capture to avoid repeated approval prompts. The Python helper takes the full screen, a pixel region or the frontmost window, and on macOS also every window of an app matched by name, one window by title or id, and a list of window ids. On Linux it uses the first of scrot, gnome-screenshot or ImageMagick's import that it finds. A PowerShell helper covers Windows, and plain screencapture, scrot, gnome-screenshot and import commands are the fallback. On macOS each display or matching window gets its own file; the agent views them in order and always reports where it saved them. One example workflow captures a running app to compare it with its Figma design. The openai/skills repository is marked deprecated; its README points to openai/plugins for current Codex skills.

Summarizes the skill at 49f948f.