Markdown を読むための canvas を GitHub Copilot app に追加してみたThe author ported the macOS Markdown reader SkimDown to Windows via WinUI 3 +…

Reading long technical documents such as design specifications and requirements written in Markdown is a routine part of software work, yet comfortable tooling for it remains fragmented across operating systems. A recent post on the Zenn blogging platform documents one developer's path from porting a macOS Markdown reader to Windows to embedding that same reading experience inside the GitHub Copilot app as a canvas extension, an example worth noting because it shows how AI assistant surfaces are being extended with custom viewers.

The origin of the project is SkimDown, a macOS Markdown reader built by the author's colleague, referred to as Tsuchida-san. According to the post, SkimDown is not a general-purpose editor but a viewer tuned specifically for reading long Markdown files, the kind used for design documents and specifications, in a fast, low-friction way. The author expresses a personal liking for the app, which set the stage for the work that followed.

After obtaining permission from the original creator, the author built a Windows version called SkimDown for Windows using WinUI 3 and WebView2. WinUI 3 is Microsoft's native user interface framework, distributed as part of the Windows App SDK and intended for building modern desktop applications on Windows. WebView2 is a Chromium-based control that lets developers embed web content inside native applications. Because Markdown is commonly rendered to HTML, pairing a native shell with an embedded web view is a practical approach for displaying formatted documents while keeping the surrounding application native.

Up to that point, the author frames the effort as a straightforward desktop application port. The more distinctive step came from combining it with separate work the author had been doing on canvas extensions for the GitHub Copilot app. Rather than leaving SkimDown for Windows as a standalone program, the author integrated the Markdown reading capability as a canvas extension, so that long documents can be read directly within the Copilot app environment.

Canvas, in this context, appears to be an extension surface within the GitHub Copilot app where custom content and interactive views can be presented alongside the assistant. The idea of a canvas as an interactive workspace has become more common across AI tools, and the author's approach of packaging a document viewer as such an extension fits that broader pattern. By reusing the WebView2-based rendering already built for the Windows port, the reader can likely be surfaced inside Copilot without rebuilding the viewing logic from scratch.

For context, GitHub Copilot began as an AI pair-programming tool focused on code completion and has expanded over time into chat, an agent-style app, and various integrations across editors and the GitHub platform. Extending the app with custom canvas experiences reflects a wider industry trend in which AI assistants are moving from single-purpose chat boxes toward hosting embeddable panels and mini-applications. Comparable concepts include interactive canvases and artifact-style panels found in other AI assistants, which similarly aim to render richer content than plain conversational text.

The technical throughline of the project is reuse. The same rendering foundation, Markdown converted to HTML and displayed through WebView2, serves both the standalone Windows application and the Copilot canvas extension. This is a pragmatic pattern for developers who want to target multiple hosts, since the presentation layer can remain consistent while the surrounding container changes from a native window to an embedded assistant surface.

As a personal project shared through a blog, the post is best read as a practical account rather than an official product announcement, and details such as availability, licensing, and long-term maintenance are not the