Windows Desktop Automation Architect

About this skill

What it does

Given a description of a repetitive Windows task (or a screenshot of the app), the skill produces a working automation plan and script: it picks the right selector strategy per UI element, sequences hotkeys and window waits correctly, adds OCR and image-matching fallbacks for elements UI Automation can't see, wraps every action in retry/timeout logic, and logs enough detail that a failure can be diagnosed without re-running the whole flow.

What is included

• Full SKILL.md with the selector-strategy decision tree (UI Automation > image match > OCR)
• Reference patterns for window waits, focus handling, and dialog interrupts
• Retry, timeout, and failure-screenshot logging template
• Guidance for Citrix/Remote Desktop sessions where accessibility trees are unreliable
• Worked example: automating a data-entry task in a legacy Windows line-of-business app

How to use it

1. Download the windows-desktop-automation-architect-SKILL.md file
2. Drop it into your agent's skills folder (e.g. .claude/skills/ for Claude Code, or upload directly in a Claude.ai chat)
3. Describe the Windows task you want automated, ideally with a screenshot of the target app
4. Review the generated design note and script, then run it in a test environment before production use

Examples

I need to automate exporting a daily report from an internal Windows accounting app (no API) and saving it to a shared drive. The app is old, some dialogs pop up with a delay, and I'm running it over Remote Desktop.

Design note: using UI Automation as the primary selector layer since the app exposes standard Win32 controls; falling back to image matching for the custom-drawn export button UI Automation can't see; adding a 2-second dialog-wait loop before each click since RDP introduces latency; wrapping the save step in a 3-retry block with a screenshot-on-failure logger. Followed by a runnable pywinauto script implementing exactly that plan.

FAQ