win-mouse-native

What This Skill Does

The win-mouse-native skill provides OpenClaw with low-level control over the Windows mouse cursor. By leveraging Windows user32.dll through a set of PowerShell and CMD scripts, this skill enables programmatic movement, clicking, and dragging of the mouse pointer. It provides a deterministic way for the AI agent to interact with desktop applications that do not offer specific APIs or screen-reading capabilities, effectively turning the agent into a hands-on desktop operator.

Installation

Installation is handled via the ClawHub distribution system. To get started, execute clawhub install openclaw/skills/skills/lurklight/win-mouse-native. After installation, you must ensure the package scripts are correctly placed in your execution path. Save the provided win-mouse.cmd.txt as win-mouse.cmd and the scripts/win-mouse.ps1.txt as scripts/win-mouse.ps1. Ensure your PowerShell execution policy allows for script execution to enable the agent to interface with the Windows kernel-level inputs.

Use Cases

This skill is essential for automating repetitive tasks that lack keyboard shortcuts or accessible menu items. Typical use cases include clicking specific buttons in legacy desktop software, automating UI testing for Windows applications, performing drag-and-drop operations between windows, or interacting with popup dialogs that require manual confirmation. It is particularly useful when the AI agent needs to move from a CLI-based workflow to an interactive GUI-based environment.

Example Prompts

1. "Click the Save button located in the active window menu bar."
2. "Move the mouse 50 pixels to the right and 10 pixels down from its current position to hover over the status icon."
3. "Perform a right-click on the desktop background to open the context menu."

Tips & Limitations

This skill is strictly for Windows environments; it will not function on Linux or macOS. Because it uses SetCursorPos and SendInput, it interacts directly with the system's input stream. Always use relative movements (dx, dy) when possible to prevent coordinate mismatch errors across different screen resolutions. For high-precision tasks, verify the current screen scaling settings, as 100% scaling is recommended for predictable coordinate calculation. When in doubt, start with small, incremental movements to verify positioning before executing primary clicks. Note that this skill simulates hardware input, meaning that if a user manually moves the mouse while the script is executing, the results may be unpredictable.