OpenAI Screenshot Skill

The OpenAI Screenshot Skill is an official, deliberately narrow workflow guide from the curated openai/skills repository. It is intended for Codex and compatible coding agents when a person explicitly asks for a desktop or system screenshot, a specific application or window, or a pixel region. According to the provider, the agent should first check whether the tool already being used has an integrated capture capability. Only when that option is unavailable, or when the request specifically calls for an operating-system capture, should this skill provide the general capture workflow. That keeps a Figma, browser, or Electron capture inside a specialized tool instead of unnecessarily exposing the whole desktop.

Purpose and triggers

The skill does not describe image editing, automatic visual assessment, or a permanent screenshot service. It gives rules for deciding when a capture is appropriate and where the file should be stored. A destination path supplied by the user takes priority. If no path is supplied, the capture should go to the operating system's normal screenshot location. When Codex needs an image only for its own inspection, the source recommends a temporary directory. This distinction prevents review artifacts from being left accidentally in a project, commit, or shared output directory.

Tool selection on Linux and macOS

On macOS, the source describes a screen-recording permission preflight before capturing applications or windows. A bundled helper can then capture the full screen, an application, a window, a window by identifier, or a rectangular region. On Linux, the supplied Python helper automatically selects the first available tool from scrot, gnome-screenshot, and ImageMagick import, according to the provider. Full-screen captures use the virtual desktop. Region captures require scrot or ImageMagick. Application and window names are documented as macOS-specific; on Linux, an active window or a known window identifier is the alternative. On Windows, the source uses a PowerShell helper with comparable options for the default location, temporary output, a region, and the active window.

Reproducible and safe use

A screenshot can expose much more than the requested application. Notifications, private messages, access hints, filenames, customer data, and tokens may appear at the edge of the image. Before taking a full-screen capture, inspect the visible desktop, close windows that are not needed, and restrict the region when that is sufficient. Do not place secrets in a screenshot, prompt, ticket, conversation, or repository. A capture is a copy of potentially confidential information and must be handled accordingly. Human review belongs in the workflow before an image is published or shared, even when the agent created the file correctly.

Boundaries and positioning

The skill depends on available operating-system tools and the required permissions. It does not replace privacy review or approval to share screen images. It is also not a substitute for specialized capture functions in Playwright, browser tooling, Figma, or Electron. According to the provider, the integrated capture capability should be preferred because it can be more targeted and reproducible. The skill is useful for development and documentation when a narrowly scoped visual proof is needed, such as a bug state, a layout comparison, or a particular dialog. It does not guarantee that what appears on screen is correct, complete, or legally shareable. Current Codex capabilities and account terms should be checked directly with OpenAI; this catalog entry therefore states no concrete prices.