I hope you've turned off VS Code's "workspace trust" settings.

https://code.visualstudio.com/docs/editor/workspace-trust

Sometimes but I wonder to what degree it actually matters. Tasks, debuggers, extensions etc. run in the context of the VM, not the host. The Remote SSH extension turns VS Code into a "thin" client which presents pretty much just the UI.

https://code.visualstudio.com/docs/remote/ssh

Readme says: https://marketplace.visualstudio.com/items?itemName=ms-vscod...

> A compromised remote could use the VS Code Remote connection to execute code on your local machine.

So I would say that it might be a bit harder for an attacker to gain access to your local machine, but you should not rely on it, because it's more like security by obscurity.

Well damn. I was under the impression that the communication channel uses/accepts only well defined VSCode specific messages related to the UI...

Darn. Maybe the solution is to use vs-code client in the browser? Like vscode.dev or https://github.com/coder/code-server ? It limits what keyboard shortcuts and extensions are available, but at least it's in a secure sandbox on the client side.

With Qubes, you can do GPU passthrough: https://forum.qubes-os.org/t/another-2-gpu-passthrough-post/....