computer-use

Automate desktop GUI interactions via background clicking, typing, and scrolling across macOS, Windows, and Linux.

16|Updated Sep 22, 2026
One-click install
npx skills add https://github.com/igniteenow/robo --skill computer-use-igniteenow
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: computer-use
Source: https://github.com/igniteenow/robo/tree/main/skills/autonomous-ai-agents/computer-use
Command: npx skills add https://github.com/igniteenow/robo --skill computer-use-igniteenow

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Automating desktop applications normally requires stealing the user's cursor and keyboard focus, interrupting their work. This Skill drives native GUI apps in the background — clicking, typing, scrolling, and dragging — without moving the real cursor, switching virtual desktops, or popping windows to the front. ## Core Features & Use Cases - Element-indexed interaction: Capture screenshots with numbered overlays on every interactable element plus an accessibility-tree index, then click by element index instead of unreliable pixel coordinates. - Verify-and-escalate ladder: Every input action returns a structured verdict (confirmed, unverifiable, suspected_noop) with recommended escalation from background element clicks to pixel coordinates, typed browser-page routes, and finally foreground delivery. - Cross-platform support: Works on macOS (AX), Windows (UIA), and Linux (AT-SPI) with any tool-capable model, plus namespaced typed actions for supported GUI browsers. - Use Case: Ask the agent to fill out a form in a native desktop app or navigate a Figma canvas while you keep typing in your editor — the agent acts in the background and reports verified results. ## Quick Start Ask the agent to open a specific desktop application and complete a task in it using the computer_use tool, starting with a capture of that app.

Frequently Asked Questions about computer-use

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate desktop apps without taking over the mouse?▼

Use the computer_use tool, which delivers clicks, typing, and scrolling in the background through cua-driver. The real OS cursor never moves and keyboard focus is never stolen, so the user can keep working in another window while automation runs.

How to click UI elements reliably with an AI agent?▼

Capture the screen with mode="som" to get numbered overlays on every interactable element, then call click with the element index. Element indices are far more reliable than pixel coordinates for most models, and indices must be refreshed after each capture.

Does computer-use work on Windows, macOS, and Linux?▼

Yes, it is cross-platform. It uses the host accessibility framework — AX on macOS, UIA on Windows, and AT-SPI on Linux — and works with any tool-capable model including Claude, GPT, Gemini, or local OpenAI-compatible endpoints.

What should I do when a click has no effect?▼

Read the structured verdict returned by the action. If effect is suspected_noop or a refusal recommends escalation, climb the ladder: retry by pixel coordinate, use the typed browser-page route when available, then re-issue with delivery_mode="foreground" as a last resort.

When should I not use computer-use for automation?▼

Avoid it for web automation that headless browser tools can handle, for file edits better done with read_file or write_file, and for shell commands that belong in a terminal tool. Reserve it for tasks requiring the user's actual native GUI applications.

Why do captures return empty or no on-screen window?▼

On Linux, DISPLAY may be unset or the session may be pure Wayland; on Windows, you may be in Session 0 instead of the interactive desktop. Ask the user to run robo computer-use doctor, which prints a structured per-check diagnostic matrix.