Skip to content

[Feature]: Windows automation skills - bridging the platform gap #38799

@aloth

Description

@aloth

[Feature]: Windows automation skills - bridging the platform gap

Problem

OpenClaw ships 10+ macOS-specific skills (apple-notes, apple-reminders, things-mac, imsg, peekaboo) but has zero Windows-equivalent skills. Given that the recommended Windows setup is WSL2, there's a natural bridge available: WSL2 can call powershell.exe directly, enabling native Windows automation from the Linux gateway.

Several open issues highlight Windows pain points (#23178, #26160, #35807, #16821), but there's no coordinated effort to bring Windows-native capabilities to agents.

Context

I've published PowerSkills (MIT) - a PowerShell toolkit that gives agents structured access to:

  • Outlook - inbox, search, send, calendar events (COM automation)
  • Edge browser - tabs, navigation, screenshots, JS eval (CDP)
  • Desktop - screenshots, window management, keystrokes (Win32 API)
  • System - shell commands, processes, system info, clipboard

All commands return JSON with a consistent envelope (status, exit_code, data, timestamp), making them straightforward for agents to consume. It's also available on ClawHub (clawhub install powerskills).

How it works from WSL2

Since WSL2 has Windows interop, calling native PowerShell is a one-liner:

powershell.exe -ExecutionPolicy Bypass -File "/mnt/c/PowerSkills/powerskills.ps1" outlook inbox --limit 5

This means a skill running inside the Linux gateway can automate Windows-native apps (Outlook, Edge, desktop) without leaving WSL2.

Questions for maintainers

  1. Is there interest in bundled Windows skills? Or should Windows automation stay on ClawHub only?
  2. Skill gating approach - for a bundled skill, would you prefer:
    • Binary gate: "requires": {"anyBins": ["powershell.exe", "pwsh"]} (activates wherever PowerShell exists)
    • OS gate: "os": ["win32", "linux"] + binary check
    • No gate (ClawHub-only, user installs intentionally)
  3. Would a "Windows skill pack" PR be welcome? Or would you prefer individual skill PRs (one for Outlook, one for browser, etc.)?

Why this matters

  • Windows is the most common desktop OS, yet the skill ecosystem is macOS-heavy
  • WSL2 is the recommended setup, and interop works out of the box
  • Outlook/Edge/Desktop automation are high-value for knowledge workers
  • The structured JSON output pattern could serve as a template for future platform skills

Happy to contribute a PR in whatever form makes sense.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions