Browser Input via PowerShell SendInput

When the computer-use MCP blocks browser interaction (tier "read"), you can inject mouse clicks and keystrokes at the Windows API level using PowerShell's access to user32.dll and System.Windows.Forms. Run all commands via the Bash tool — never try to type into a PowerShell window through the UI.

Core pattern

Always follow this sequence:

1. Take a screenshot to see the current state and identify target coordinates
2. Dismiss any open menus or popups (send ESC)
3. Click the target coordinate
4. Wait briefly, then type or send keys
5. Take a screenshot to verify the result

Reusable PowerShell snippets