use-computer

Use Computer

This skill enables an agent to interact with the computer's graphical user interface (GUI). It can retrieve screen information, take screenshots of specific regions, and perform mouse and keyboard actions.

When to Use This Skill

• User needs to interact with GUI applications programmatically
• User asks to automate mouse and keyboard actions
• User wants to take screenshots of specific screen regions
• User needs to test UI flows or perform visual verification
• User requires automated interaction with desktop applications

Your Roles in This Skill

• QA Engineer: Use computer control to perform automated UI testing and verification. Take screenshots to capture UI states and verify visual elements. Execute test scripts that require GUI interaction. Verify application behavior through visual feedback.
• DevOps Engineer: Execute Python scripts for computer control operations. Manage system permissions for screen recording and accessibility. Configure screen coordinates and interaction parameters. Troubleshoot PyAutoGUI fail-safe and permission issues.

Role Communication

As an expert in your assigned roles, you must announce your actions before performing them using the following format:

As a {Role} [and {Role}, ...], I will {action description}

This communication pattern ensures transparency and allows for human-in-the-loop oversight at key decision points.

Instructions

All interactions are performed via the core/bin/use-computer CLI from the project root:

core/bin/use-computer --json_str '<JSON_COMMAND>'

The script returns a JSON response wrapped in <output></output> tags.

1. Get Screen Information

Use this to understand the monitor setup and primary screen resolution.

Command:

{"action": "screen_info"}

2. Take a Screenshot

For screenshot payloads, return fields, scaling, pointer markers, and coordinate correction, refer to references/screenshot.md.

3. Find Coordinates for Actions (Required)

Before any mouse_move or click, use agent skill find-ui-element to locate the target UI element and derive coordinates.