Desktop Control Skill

🤖 AI Agent Skill for desktop automation using PyAutoGUI.

Control mouse, keyboard, and screen programmatically through a simple CLI interface.

⚡ Quick Install

For AI Agents

Add this skill to your AI coding agent with a single command:

npx skills add patrickporto/desktop-agent

For Users

Install the CLI with pipx (recommended):

pipx install desktop-agent

Or run without installing using uvx:

uvx desktop-agent --help

Or using pip:

pip install desktop-agent

---

🎯 For AI Agents

This project is packaged as an AI Agent Skill. To use it:

1. Read the skill documentation: SKILL.md
2. Install the CLI: Use pip install desktop-agent or pipx install desktop-agent
3. Invoke commands: Use desktop-agent <category> <command>

Quick Reference for Agents:

• All commands documented in SKILL.md
• Practical examples in examples/automation_examples.md
• Help system: desktop-agent --help
• All commands return JSON output by default

---

📦 Installation

Using pipx (Recommended)

pipx install desktop-agent

Using pip

pip install desktop-agent

Run without installing (using uvx)

uvx desktop-agent

Usage

The CLI is organized into command categories:

🖱️ Mouse (mouse)

# Move mouse to coordinates
desktop-agent mouse move 100 200

# Move with duration (animation)
desktop-agent mouse move 100 200 --duration 1.0

# Click at current position
desktop-agent mouse click

# Click at specific coordinates
desktop-agent mouse click 500 500

# Right click
desktop-agent mouse right-click

# Double click
desktop-agent mouse double-click 300 400

# Drag to coordinates
desktop-agent mouse drag 200 300

# Scroll (positive = up, negative = down)
desktop-agent mouse scroll 5
desktop-agent mouse scroll -3

# Get current mouse position
desktop-agent mouse position

⌨️ Keyboard (keyboard)

# Write text
desktop-agent keyboard write "Hello World"

# Write with interval between keys
desktop-agent keyboard write "Slow typing" --interval 0.1

# Press a key
desktop-agent keyboard press enter

# Press multiple times
desktop-agent keyboard press a --presses 5

# Execute keyboard shortcut
desktop-agent keyboard hotkey "ctrl,c"
desktop-agent keyboard hotkey "ctrl,shift,esc"

# Hold/release key
desktop-agent keyboard keydown shift
desktop-agent keyboard keyup shift

🖼️ Screen (screen)

# Capture screenshot (full screen)
desktop-agent screen screenshot my_screen.png

# Take screenshot of active window
desktop-agent screen screenshot active_window.png --active

# Take screenshot of specific window
desktop-agent screen screenshot notepad.png --window "Notepad"

# Screenshot of specific region (x,y,width,height)
desktop-agent screen screenshot region.png --region "100,100,500,400"

# Locate image within active window
desktop-agent screen locate button.png --active

# Locate center of image on screen
desktop-agent screen locate-center button.png --confidence 0.8

# Find text coordinates within active window
desktop-agent screen locate-text-coordinates "OK" --active

# Find text in specific image
desktop-agent screen locate-text-coordinates "Confirm" --image screenshot.png

# Case-sensitive search
desktop-agent screen locate-text-coordinates "Login" --case-sensitive

# Read all text from screen
desktop-agent screen read-all-text

# Read text from image
desktop-agent screen read-all-text --image capture.png

# Specify languages for OCR (default: pt,en)
desktop-agent screen locate-text-coordinates "Button" --lang "en"

💬 Messages (message)

# Show alert
desktop-agent message alert "Hello!"

# Confirmation
desktop-agent message confirm "Are you sure?"

# Input prompt
desktop-agent message prompt "Enter your name:"

# Password
desktop-agent message password "Enter your password:"

📱 Applications (app)

# Open an application (cross-platform)
desktop-agent app open notepad
desktop-agent app open "Google Chrome"

# Open with arguments
desktop-agent app open chrome --arg "https://google.com"

# Focus on a window by title
desktop-agent app focus "Untitled - Notepad"

# List all visible windows
desktop-agent app list

Automation Examples

Open Notepad and write

desktop-agent app open notepad
desktop-agent app focus notepad
desktop-agent keyboard write "Hello from Desktop Skill!"

Capture screenshot and analyze

desktop-agent screen screenshot full_screen.png
desktop-agent screen pixel 500 500

Available Commands

Run desktop-agent --help to see all commands:

desktop-agent --help
desktop-agent mouse --help
desktop-agent keyboard --help
desktop-agent screen --help
desktop-agent message --help

Project Structure

desktop-skill/
├── desktop_agent/       # Main package
│   ├── __init__.py
│   ├── commands/        # Command modules
│   │   ├── __init__.py
│   │   ├── mouse.py    # Mouse commands
│   │   ├── keyboard.py # Keyboard commands
│   │   ├── screen.py   # Screen/screenshot/OCR commands
│   │   └── message.py  # Message boxes
├── pyproject.toml      # Project configuration
└── README.md           # This documentation

Technologies

• PyAutoGUI: GUI automation
• EasyOCR: Optical character recognition
• Typer: Modern CLI framework