Skip to content
Skills Library
Programming · Review & audit

screencapture

/screencapture

Captures the user's macOS screen, a window, or a region with the built-in `screencapture` CLI, then reads the image so the agent can see what the user sees. Use it when the user asks to check how a native macOS app looks (SwiftUI, AppKit, Electron, Qt, Tk, Java Swing), verify a window rendered correctly after a build or code change, diagnose a visual bug they describe in a desktop app, confirm a menu, dialog, or layout state, or asks "can you see this?" about anything outside a browser, even if they don't say "screenshot". Web pages and web apps belong to a browser automation tool.

Repository
skills ↗
Licence
MIT
Last repository change
12.10.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 12.10.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Captures the user's macOS screen, a window, or a region with the built-in `screencapture` CLI, then reads the image so the agent can see what the user sees. Use it when the user asks to check how a native macOS app looks (SwiftUI, AppKit, Electron, Qt, Tk, Java Swing), verify a window rendered correctly after a build or code change, diagnose a visual bug they describe in a desktop app, confirm a menu, dialog, or layout state, or asks "can you see this?" about anything outside a browser, even if they don't say "screenshot". Web pages and web apps belong to a browser automation tool.

Triggers from the description

can you see this?t say
Original author description

Captures the user's macOS screen, a window, or a region with the built-in `screencapture` CLI, then reads the image so the agent can see what the user sees. Use it when the user asks to check how a native macOS app looks (SwiftUI, AppKit, Electron, Qt, Tk, Java Swing), verify a window rendered correctly after a build or code change, diagnose a visual bug they describe in a desktop app, confirm a menu, dialog, or layout state, or asks "can you see this?" about anything outside a browser, even if they don't say "screenshot". Web pages and web apps belong to a browser automation tool.

How it works

  1. Core workflow
  2. Picking the right mode
  3. Full flag reference
  4. Cleanup is mandatory
  5. When NOT to use this skill
Original SKILL.md headings · 96 lines, 955 words
SKILL.md preview
---
name: screencapture
description: Captures the user's macOS screen, a window, or a region with the built-in `screencapture` CLI, then reads the image so the agent can see what the user sees. Use it when the user asks to check how a native macOS app looks (SwiftUI, AppKit, Electron, Qt, Tk, Java Swing), verify a window rendered correctly after a build or code change, diagnose a visual bug they describe in a desktop app, confirm a menu, dialog, or layout state, or asks "can you see this?" about anything outside a browser, even if they don't say "screenshot". Web pages and web apps belong to a browser automation tool.
---

# screencapture

macOS ships `screencapture`, a fast CLI for grabbing the screen. Use it to _see_ what the user sees when working on native GUI software. Web apps belong in a browser-automation tool; this skill is for everything else (SwiftUI/AppKit, Electron, Java/Swing, Qt, Tk, games, system UI, Finder, Xcode, etc.).

## Core workflow

1. **Capture** to the scratch directory with a unique filename. Pick the directory in this order: the session scratch directory when the host names one in its instructions (Claude Code calls it the scratchpad directory), else `$TMPDIR`, else `/tmp`. A session directory is isolated from other sessions and from the user's project, so files there cannot collide or clutter shared space. Set `OUT` to that path and use it in every command below.
2. **Read** the image with the Read tool — Agent is multimodal and will see the PNG inline.
3. **Delete** the file immediately after reading. The user wants temp space kept clean; never leave screenshots lying around.

Filename convention: `$OUT/screencap-$(date +%s).png`. The timestamp prevents collisions if you capture multiple times in one task.

## Picking the right mode

| Goal                                                             | Command                                                                                   |
| ---------------------------------------------------------------- | ----------------------------------------------------------------------------------------- |
| Whole main display, no UI sound                                  | `screencapture -x -m $OUT/screencap-<ts>.png`                                             |
| All displays                                                     | `screencapture -x $OUT/screencap-<ts>.png` (one file per screen — macOS appends suffixes) |
| User picks a window or region interactively                      | `screencapture -x -i $OUT/screencap-<ts>.png`                                             |
| Specific rectangle (x,y,width,height in points, origin top-left) | `screencapture -x -R 100,100,800,600 $OUT/screencap-<ts>.png`                             |
| Specific window by id                                            | `screencapture -x -l <windowid> $OUT/screencap-<ts>.png`                                  |
| Capture after delay (e.g. to let a menu open)                    | `screencapture -x -T 3 $OUT/screencap-<ts>.png`                                           |

Always pass `-x` so the shutter sound doesn't startle the user.

`-i` is interactive: it blocks until the user clicks/drags. Only use it when you genuinely need the user to point at something — for autonomous "show me the app's main window" checks, prefer full-screen (`-m`) or a known rect (`-R`).

To find a window id for `-l`, you can run `osascript -e 'tell app "System Events" to get id of windows of process "<AppName>"'` or use `CGWindowListCopyWindowInfo` via a small helper; this is rarely worth it — full-screen capture is usually fine.

## Full flag reference

```text
usage: screencapture [-icMPmwsWxSCUtoa] [files]
  -c         force screen capture to go to the clipboard
  -b         capture Touch Bar — non-interactive modes only
  -C         capture the cursor as well as the screen. only in non-interactive modes
  -d         display errors to the user graphically
  -i         capture screen interactively, by selection or window
               control key — causes screenshot to go to clipboard
               space key   — toggle between mouse selection and window selection modes
               escape key  — cancels interactive screenshot
  -m         only capture the main monitor, undefined if -i is set
  -D<display> screen capture or record from the display specified. -D 1 is main display, -D 2 secondary, etc.
  -o         in window capture mode, do not capture the shadow of the window
  -p         screen capture will use the default settings for capture. The files argument will be ignored
  -M         screen capture output will go to a new Mail message
  -P         screen capture output will open in Preview or QuickTime Player if video
  -B<bundleid> screen capture output will open in app with bundleid
  -s         only allow mouse selection mode
  -S         in window capture mode, capture the screen not the window
  -J<style>  sets the starting of interactive capture
               selection       - captures screen in selection mode
               window          - captures screen in window mode
               video           - records screen in selection mode
  -t<format> image format to create, default is png (other options include pdf, jpg, tiff and other formats)
  -T<seconds> take the picture after a delay of <seconds>, default is 5
  -w         only allow window selection mode
  -W         start interaction in window selection mode
  -x         do not play sounds
  -a         do not include windows attached to selected windows
  -r         do not add dpi meta data to image
  -l<windowid> capture this windowsid
  -R<x,y,w,h> capture screen rect
  -v         capture video recording of the screen
  -V<seconds> limits video capture to specified seconds
  -g         captures audio during a video recording using default input.
  -G<id>     captures audio during a video recording using audio id specified.
  -k         show clicks in video recording mode
  -U         Show interactive toolbar in interactive mode
  -u         present UI after screencapture is complete. files passed to command line will be ignored
  -H         capture content in HDR
  files      where to save the screen capture, 1 file per screen
```

## Cleanup is mandatory

The user keeps temp space tidy. After every `Read` of the screenshot, immediately `rm` it in the same logical step. Don't batch — capture, read, delete, one screenshot at a time. If you need a sequence of screenshots, still delete each one as soon as you've consumed it.

Example flow:

```bash
OUT="${SCRATCH_DIR:-${TMPDIR:-/tmp}}"   # SCRATCH_DIR = the session scratch path from the host's instructions, if any
screencapture -x -m "$OUT/screencap-1715300000.png"
# → Read tool on $OUT/screencap-1715300000.png
rm "$OUT/screencap-1715300000.png"
```

If a capture fails (e.g. no Screen Recording TCC permission), `screencapture` writes a black or empty image rather than erroring loudly. If the Read shows a black/empty frame, tell the user to grant Screen Recording permission in System Settings → Privacy & Security → Screen Recording for the terminal app running Agent, then retry.

## When NOT to use this skill

- Web pages, web apps, anything in a browser → use the browser automation skill.
- The user asks for a screen _recording_ (video) → `screencapture -v` works but is rarely what's wanted; ask first.
- You just want to know what app is focused → `osascript -e 'tell app "System Events" to name of first process whose frontmost is true'` is faster and doesn't need TCC.

Versions

The database has no other skills named screencapture.

Database snapshot: 13.11.2026 · package version a06d7874e64d (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100screencapture-mac
    Captures the user's macOS screen, a window, or a region with the built-in `screencapture` CLI, then reads the image so the agent can see what the user sees. Use it when the user asks to check how a native macOS app looks (SwiftUI, AppKit, Electron, Qt, Tk, Java Swing), verify a…
  • 95capture-screen
    Programmatic screenshot capture on macOS. Find window IDs with Swift CGWindowListCopyWindowInfo, control application windows via AppleScript (zoom, scroll, select), and capture with screencapture. Use when automating screenshots, capturing application windows for documentation,…
  • 90desktop-swiftui
    Use when the user asks about SwiftUI for macOS, SwiftUI desktop patterns, AppKit bridging, SwiftUI app lifecycle, or SwiftUI multiplatform apps. Do NOT use for: AppKit-only (desktop-appkit), or iOS SwiftUI (mobile skill).
  • 90desktop-appkit
    Use when the user asks about macOS app development with AppKit, Cocoa, Objective-C, Swift AppKit, macOS UI framework, or native Mac application architecture. Do NOT use for: SwiftUI (desktop-swiftui), or cross-platform desktop (desktop-electron, desktop-tauri).
  • 100appkit-app-architecture-workflow
    Guide AppKit app-structure decisions for macOS apps across app delegates, status-item menu bar apps, responder-chain menus, windows, controllers, restoration, archiving, Observation, and mixed AppKit/SwiftUI composition. Use when the user wants help deciding where AppKit…
  • 100build-and-verify
    Build a SwiftUI macOS app from the command line, install it, and check light and dark window snapshots of every page before calling a change done. Use after any UI code change, when a build fails, or when the user asks to run, screenshot, or check the app.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.