· · 2 minutes read

A WinUI Agent Plugin for GitHub Copilot and Claude Code

Microsoft released agent skills for WinUI development — scaffold, build, run, test, iterate, all with GitHub Copilot CLI or Claude Code. The key innovation: purpose-built tools that ground the agent in WinUI-specific facts.

WinUI Windows App SDK GitHub Copilot AI Agents
This post is also available in:Català, Español, Deutsch, Français, Português, Italiano, 日本語, 中文, 한국어, Русский, हिन्दी, Polski, Türkçe, العربية, Bahasa Indonesia, Nederlands

Getting AI agents to write WinUI apps has been hit-or-miss. Generic agents pull from too much context — mixing WinUI 3 with older UWP patterns, getting MSIX packaging wrong, or generating apps that don’t actually run. The new WinUI agent plugin takes a different approach.

What It Is

A set of open-source agent skills designed specifically for WinUI and Windows App SDK development, compatible with both GitHub Copilot CLI and Claude Code.

Install and run:

# Add the WinUI plugin to GitHub Copilot CLI
/plugin install winui@awesome-copilot

# Run the setup skill to install prerequisites
/winui:winui-setup

Then describe the app you want. The agent handles the rest.

The End-to-End Loop

The skills are built around a complete development loop: scaffold → build → run → test → iterate. What makes this different from just asking a generic agent:

Scaffold correctly — the agent picks the right dotnet new WinUI template, not an older UWP pattern.

Build through the right pipeline — WinUI apps need packaged execution. Generic agents often don’t know this and generate apps that compile but don’t run. The skills handle the packaged execution model correctly.

Interact and validate — if you ask, the agent can interact with the running app to validate functionality, not just check that the build succeeded.

Fix compilation errors automatically — rather than stopping at the first error, the agent recognizes common WinUI failure modes and steers toward working solutions.

Beyond the basics, the skills also handle higher-level polish: accessibility audits, code reviews, MSIX packaging.

The Token Efficiency Design

Building in rich WinUI context up front would make every turn expensive. Instead, the skills are paired with purpose-built tools that give the agent ground-truth answers on demand — only pulling in specific reference data when the agent needs it, not loading entire documentation pages preemptively.

The result: agents that know WinUI deeply without the token overhead of loading everything every turn.

Supported Features

The skills handle: XAML and Fluent Design, MVVM, MSIX packaging, code signing, Store submission, accessibility, theming, UI automation. That covers most of the surface area where agents typically get stuck.

Original post: Introducing WinUI agent plugin for GitHub Copilot and Claude Code

Share:
View source code for this post on GitHub ↗
← VS Code 1.121: Pin Favorite Models, Terminal Output Compression, Agent SSH