macOS Computer Use

Use this skill when you need to control a macOS app by looking at screenshots and sending real input events like a user.

Before You Start

• If something does not work, ask the user to check the main app that is running the agent.
• Do not guess that the main app is a terminal.
• The main app might be Cursor, an IDE, a plugin host, a desktop app, or something else.
• That main app needs Accessibility permission for mouse moves, clicks, typing, and scrolling.
• That same app may also need Screen & System Audio Recording permission for screenshots.

Simple rule:

• If clicking or typing fails, ask the user to check Accessibility for the main app.
• If screenshots fail, ask the user to check Screen & System Audio Recording for the main app.

Main Idea