screenshot
by openaiThe screenshot skill helps capture a full screen, app window, or pixel region when you need an OS-level image instead of a browser-only capture. Use it for screenshot usage in Workflow Automation, with save-location rules, macOS permission handling, and clear install guidance for reliable desktop captures.
This skill scores 78/100, which means it is a solid listing candidate for directory users who need reliable screenshot capture guidance. The repository clearly defines when to use it, provides operational scripts for macOS and cross-platform capture, and includes preflight steps that reduce guesswork, though the install decision would still benefit from a clearer quick-start and explicit install command.
- Explicit trigger guidance: use when a user asks for a desktop/system screenshot or when tool-specific capture is unavailable.
- Operational depth: includes dedicated scripts for macOS, Windows, and Python-based capture, plus macOS permission preflight to reduce repeated prompts.
- Good agent leverage: the skill specifies save-location rules and tool priority, helping an agent choose the right capture path with less ambiguity.
- No install command in SKILL.md, so adopters may need to infer setup rather than follow a one-step install flow.
- The excerpt is strong on mechanics but still leaves some platform-specific execution details to the scripts, which may require inspection for edge cases.
Overview of screenshot skill
What the screenshot skill does
The screenshot skill helps an agent capture the right desktop image when a task needs a whole screen, a specific window, an app region, or a saved file path. It is the right fit when you need a real OS-level screenshot rather than a browser-only capture, a design-tool capture, or a generic prompt answer.
When it is the right install
Install screenshot if your workflow includes desktop apps, multi-window review, OS UI debugging, or cases where tool-specific capture is unavailable. It is especially useful for Workflow Automation jobs that need visual proof, handoff artifacts, or pixel-accurate references.
What makes it different
This screenshot skill is decision-oriented, not just a command wrapper. It encodes capture priority, save-location rules, and macOS permission handling so the agent can choose a workable path faster and with fewer prompts. That reduces guesswork when the user only says “take a screenshot” or gives an incomplete target.
How to Use screenshot skill
Install and locate the core files
Install with npx skills add openai/skills --skill screenshot. Then read SKILL.md first, followed by scripts/take_screenshot.py, scripts/ensure_macos_permissions.sh, and agents/openai.yaml. If you need platform-specific behavior, inspect the Swift helpers in scripts/ before assuming the capture path.
Give the skill a complete capture brief
A strong screenshot usage request names four things: target, area, output path, and constraints. For example: “Capture the active Photoshop window and save it to /tmp/review.png” or “Take a full-screen screenshot of display 2 in the default screenshot folder.” If the path is omitted, the skill follows the OS default; if Codex needs the image for inspection, it should save to temp.
Use the right workflow for the platform
For browsers, Figma, or Electron, prefer their native or tool-specific screenshot path first. Use this skill when you need the whole desktop, when app/window capture is the real requirement, or when another tool cannot capture what matters. On macOS, run the permission preflight before window/app capture to avoid repeated Screen Recording prompts.
Practical prompt pattern
A good screenshot guide prompt is specific enough to trigger the correct helper and output. Example: “Use the screenshot skill to capture the left half of the editor window on macOS, then save it to the default screenshot location.” If you need a region, provide coordinates in x,y,w,h form and say whether the region is relative to the screen or to a window.
screenshot skill FAQ
Is screenshot only for full-screen captures?
No. The screenshot skill covers full-screen, window, app, and region captures. Use full-screen only when the entire desktop context matters; otherwise narrow the target to reduce noise and improve usefulness.
When should I not use this skill?
Do not use screenshot when a better-integrated tool can capture the exact surface you need, such as a Figma or browser-specific workflow. Also avoid it if your goal is text extraction or UI reasoning without needing an actual image artifact.
Do beginners need to know OS details first?
No. Beginners can use the skill with a plain request like “take a screenshot of this window.” The main improvement comes from adding the target, save path, and any crop details. On macOS, permissions may still be the main blocker, so expect one setup step.
How is this different from a generic prompt?
A generic prompt may describe the desired image, but the screenshot skill also handles capture choice, save-location rules, and macOS permission friction. That makes it more reliable for Workflow Automation because the agent is guided toward a concrete file output instead of an abstract answer.
How to Improve screenshot skill
Give the clearest target possible
The biggest quality gain comes from naming exactly what should appear in frame. “Capture the editor” is weaker than “capture the VS Code window showing app.py with the terminal visible.” Specific targets reduce failed captures, especially when multiple similar windows are open.
Add constraints that change the result
If the screenshot must exclude private content, include only one monitor, or show a specific resolution, say so up front. For region captures, provide coordinates and explain whether the crop should include chrome, title bars, or only content. These details matter more than extra prose.
Use the first output to refine the next one
If the screenshot is too broad, too small, or missing the relevant UI state, iterate by tightening the target and referencing what was wrong. For example: “Retake with only the modal visible” or “Move the crop down to include the status bar.” That feedback loop is the fastest way to improve screenshot usage.
Watch for common failure modes
The usual problems are permission prompts on macOS, capturing the wrong monitor, and asking for a screenshot when a better tool could have delivered a cleaner result. If the first capture fails, improve the request by adding the app name, window title, screen number, or exact region.