Add zoom to anthropic computer

[farrelmahaztra]

Note

Medium Risk
Touches core computer-control pathways (mouse/scroll/key dispatch) and adds new image-processing logic that may behave differently across platforms and screen scaling.

Overview
Adds a new Anthropic zoom computer action that validates a region box, scales it from agent coordinates to screen coordinates, and returns a cropped/resized screenshot via a new executor zoom() API.

Extends executor implementations (BaseExecutor, PyAutoGUIExecutor, XDOExecutor) with shared crop/resize logic and real screenshot-based region extraction, and updates Anthropic interactions to support modifier-held mouse actions (click/scroll via hold_keys) plus repeated key presses via a new repeat parameter (capped at 100, only screenshot on the final repeat). Adds a focused unit test verifying zoom coordinate scaling and executor invocation.

^{Written by Cursor Bugbot for commit 745db08. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}