PowerSkills: Giving AI Agents Control Over Windows with PowerShell

If you’re building AI agents that need to interact with Windows, you’ve likely hit a wall: most agent tooling targets Linux or macOS, leaving Windows automation as an afterthought. PowerSkills changes that.

PowerSkills is an open-source PowerShell toolkit that gives AI agents structured control over Windows – Outlook email, Edge browser, desktop windows, and system operations. Every action returns clean, parseable JSON that agents can reliably process.

Why PowerSkills?

Windows is where enterprise work happens. Outlook holds the emails, Edge holds the browser sessions, and PowerShell is the automation backbone. Yet most AI agent frameworks assume a Unix environment. PowerSkills bridges this gap with four focused skill modules:

• Outlook – Read inbox, search emails, send messages, access calendar events via COM automation
• Browser – Control Edge via Chrome DevTools Protocol (CDP) – list tabs, navigate, take screenshots, interact with the DOM
• Desktop – Manage windows, capture screenshots, read/write clipboard, send keystrokes via Win32 API
• System – Query system info, manage processes, execute commands, read environment variables

Structured JSON Output

Every PowerSkills action returns a consistent JSON envelope. No more parsing free-text output or guessing at error states:

{
  "status": "success",
  "exit_code": 0,
  "data": {
    "hostname": "WORKSTATION-01",
    "os": "Microsoft Windows 11 Pro",
    "cpu": "AMD Ryzen 9 5900X",
    "memory_gb": 32,
    "uptime": "3.04:22:15"
  },
  "timestamp": "2026-03-06T17:30:00Z"
}

This makes it trivial for AI agents to check status, extract data, and handle errors programmatically – no regex gymnastics required.

Two Ways to Run

PowerSkills supports both a dispatcher mode for centralized control and standalone mode for direct access:

# Dispatcher mode - one entry point for everything
.\powerskills.ps1 system info
.\powerskills.ps1 outlook inbox --limit 5
.\powerskills.ps1 browser tabs
.\powerskills.ps1 desktop screenshot --path C:\temp\screen.png

# Standalone mode - call skills directly
.\skills\system\system.ps1 info
.\skills\outlook\outlook.ps1 inbox --limit 5

Practical Examples

Check unread emails

.\powerskills.ps1 outlook inbox --limit 3 --unread

Returns sender, subject, date, and body preview for each message – perfect for an agent that triages your inbox.

Take a screenshot of the active window

.\powerskills.ps1 desktop screenshot --active --path C:\temp\active.png

Captures and saves a screenshot, returning the file path in JSON. Useful for visual verification workflows or automated documentation.

List open browser tabs

.\powerskills.ps1 browser tabs

Returns every open Edge tab with title and URL via the DevTools Protocol. Your agent can find specific pages, navigate, or extract content.

Execute a system command

.\powerskills.ps1 system exec --command "Get-ChildItem C:\Projects -Name"

Runs any PowerShell command and wraps the output in the standard JSON envelope. The agent gets structured results instead of raw terminal output.

Agent-Friendly by Design

Each skill includes a SKILL.md file with structured metadata – name, description, available actions, and parameters. AI agents can discover and understand capabilities without hardcoded instructions. This follows the emerging pattern of self-describing agent skills.

Getting Started

No package manager, no installer, no dependencies beyond PowerShell 5.1 and Windows 10/11:

1. Clone or download the repository
2. Set execution policy: Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned
3. Run: .\powerskills.ps1 list to see available skills
4. For browser skills: launch Edge with --remote-debugging-port=9222

That’s it. Copy the folder, run the scripts.

Open Source

PowerSkills is fully open source under the MIT license. Contributions, issues, and feature requests are welcome.

Give it a star if you find it useful. Whether you’re building agent workflows, automating your Windows workstation, or just want structured JSON output from PowerShell – PowerSkills has you covered.