Skip to main content
skill CX Codex

os-use — Claude Code skill for OS automation and screen control

os-use — a skill for Claude Code, Cursor, Codex and other platforms: captures screenshots, drives the mouse and keyboard, matches images on screen, manages windows. Runs on macOS 12+ and Windows 10+.

no license 280 installs ★ 15 updated 2026-10-11

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add zrong/skills --skill os-use -a claude-code
Open on GitHub

What os-use does

os-use gives an AI agent direct access to the desktop layer. It installs into Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and OpenCode — each platform gets its own install path, and from there the agent can capture screens, click, type, and move windows.

On macOS, window management goes through AppleScript, while pyautogui and OpenCV handle screenshots and image search. On Windows the same screen and input work runs through pyautogui, but pywinauto and pygetwindow take over window management because AppleScript has no equivalent there. Browser automation on both platforms requires Chrome DevTools MCP as a separate tool. Text recognition from screenshots depends on pytesseract and the Tesseract OCR engine — the skill does not bundle them.

The Python environment must include pyautogui, opencv-python-headless, numpy, Pillow, and pyperclip before the skill can run. On macOS, the agent needs Accessibility and Screen Recording permissions granted in System Settings — without them every screenshot and click will be blocked by the OS. On Windows, interacting with elevated windows requires administrator privileges.

What it helps with

  • capture a screenshot of a specific screen region or application window
  • locate a UI element by template image and click it programmatically
  • automate text input and keyboard shortcuts across any desktop app
  • move, resize, or close application windows without user interaction
  • read and write clipboard content inside an automation script

Example prompts for the agent

  • Take a screenshot, find the button matching login_button.png, and click it
  • Move the Calculator window to the right half of the screen and resize it to 800×600
  • Type into the active input field, press Enter, then save a screenshot of the result

How to install os-use

1Skills CLI (Vercel)

npx skills add zrong/skills --skill os-use -a claude-code

2Manually: put the files in place

~/.claude/skills/os-use/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/os-use/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about os-use

How do I install os-use in Claude Code?

Run npx skills add zrong/skills --skill os-use -a claude-code in your terminal. The skill file will be placed at ~/.claude/skills/os-use/SKILL.md.

What does the os-use skill do?

It lets the agent capture screenshots, control the mouse and keyboard, find images on screen with OpenCV, and manage application windows. Browser automation goes through Chrome DevTools MCP as a separate layer.

Does os-use work the same way on macOS and Windows?

Screenshot capture, mouse/keyboard control, and image matching work the same on both. Window management differs: macOS uses AppleScript, Windows uses pywinauto and pygetwindow. There is no Hammerspoon equivalent on Windows.

Similar skills

All skills for Automation and no-code: n8n, Zapier, Make →

From the same set2

Project sponsors

Companies that keep this analytics open