agentsclimarketplace

Visual capture

Skill mustafaakben/plugin-windows-mcp/skills/visual-capture

This skill should be used when the user asks to "take a screenshot", "capture the screen", "identify elements on screen", "get a snapshot", "extract DOM content", "find interactive elements", "analyze screen state", "what is on the screen", or needs to observe and analyze the current Windows desktop state using Screenshot or Snapshot tools.From its SKILL.md

Install
npx -y skills add mustafaakben/plugin-windows-mcp --skill visual-capture

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.1 KB, as published. Nobody here has run it

Visual Capture and Analysis

This skill covers the two observation tools — Screenshot and Snapshot — which are the foundation of every automation workflow. Before acting on the screen, always observe it first.

Screenshot vs. Snapshot: When to Use Which

CriteriaScreenshotSnapshot
SpeedFastSlower
Visual imageYesYes
Cursor positionYesYes
Active window infoYesYes
Interactive element IDsNoYes
Element coordinatesNoYes
Scrollable regionsNoYes
DOM extractionNoYes (with use_dom)

Decision rule:

  • Need to verify a result or get a quick look? -> Screenshot
  • Need to identify elements before clicking/typing? -> Snapshot
  • Need web page content/structure? -> Snapshot with use_dom=True

Screenshot

Fast desktop capture returning a visual image, cursor position, and active window information.

Takes no parameters — captures the entire desktop.

When to use:

  • Verifying the result of an action (did the click work?)
  • Checking the general state of the desktop
  • Quick visual confirmation between steps
  • Monitoring progress of a long operation

Output includes:

  • Full desktop image
  • Current cursor coordinates
  • Active (foreground) window title and bounds

Snapshot

Full state capture with interactive element identification.

Key parameter: use_dom (boolean, default: false)

When to use:

  • Before any Click, Type, or interaction — to locate targets
  • When needing to enumerate all interactive elements in a window
  • When needing to identify scrollable regions
  • When automating web browsers and needing page DOM content

Output includes:

  • Full desktop image
  • List of interactive elements, each with:
    • Unique element ID
    • Screen coordinates (x, y)
    • Element type (button, text field, link, etc.)
    • Element label/text
  • Scrollable regions with boundaries
  • DOM content (only when use_dom=True)

DOM Extraction (use_dom=True)

Enable DOM extraction when automating web browsers. This extracts the page's HTML structure, enabling:

  • Reading page content without scraping
  • Identifying elements by their HTML attributes
  • Understanding page layout and structure
  • Finding hidden or off-screen elements

Performance note: DOM extraction adds latency. Only enable when specifically needed for web automation.

Effective Analysis Patterns

Identify a Button to Click

Snapshot -> scan element list for target button
  -> note its (x, y) coordinates
  -> Click(x, y)

Find a Text Field to Fill

Snapshot -> scan element list for input/text field elements
  -> identify the field by its label or position
  -> Click(field_x, field_y) -> Type(text="value")

Verify an Action Succeeded

Perform action -> Screenshot
  -> check if expected visual change occurred
  -> if not, retry or take a different approach

Analyze a Complex Window

Snapshot -> enumerate all elements
  -> categorize by type (buttons, fields, labels)
  -> identify the target interaction area
  -> plan the sequence of actions

Extract Web Page Data

Snapshot(use_dom=True) -> read DOM content
  -> extract text, links, or structured data
  -> process the extracted information

Best Practices

  1. Start every automation with Snapshot — Establish the current state before acting
  2. Use Screenshot for verification — Cheaper than Snapshot when element IDs are not needed
  3. Re-snapshot after major changes — Window content may have shifted (new elements, dialogs, navigation)
  4. Do not cache old snapshots — Screen state changes constantly; always use fresh captures
  5. Minimize DOM extraction — Only use use_dom=True when web content analysis is specifically required
  6. Read element labels carefully — Match elements by their label text, not just position, to handle UI layout variations

Handling Common Scenarios

Multiple monitors: Screenshot and Snapshot capture the primary display. To interact with secondary monitors, specify coordinates in that monitor's range.

Overlapping windows: The Snapshot returns elements for all visible windows. Focus the target window first using App(switch) before taking a snapshot.

Dynamic content: Elements that load asynchronously may not appear immediately. Use Wait(0.5-1s) before Snapshot if content is still loading.

High DPI displays: Coordinates are in screen pixels. The tools handle DPI scaling automatically.

Additional Resources

Reference Files

For advanced capture techniques and analysis workflows:

  • references/analysis-guide.md — Advanced screen analysis, multi-window strategies, and element identification patterns

Keep looking

Skills are one crate of 608,475. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.