I feel like we’re getting a little too deep into semantics vs. intent and application. I am acutely aware you really dislike the idea of someone doing this, though.
From a purely semantic point, I could equally contend that file uploads is an OS-level function, it directly involves accessing the file system. Heck, even loading / reading in the contents of a CSV file potentially could be seen as an OS-level operation. So, the browser is already ‘going there’.
This also means though, that in terms of the AUT, if the functionality of a website includes a copy to clipboard command, then that makes it functionality that the website is offering, via the browser.
Sure, the contents of the clipboard is held at an OS-level, but the contents of files used for uploading is held by the file system, so why is one okay but not the other? As from a sandboxing perspective, this point still applies, yet Chrome explicitly allows a toggle (as does FF) to allow file system access for an extension.
But to go back to intent and application, what the user is trying to do is verify that the AUT is correctly copying data to the clipboard, that the data copied is as expected, and they may then want to use that data in other operations related to testing the very same website.
So, why is a desire to test something that is potentially part of a web site or web application such a terrible thing, if the automation tool had a way and means of doing so? As I really don’t follow why it should be out of scope, as it may be the only area the user is unable to cover from the test suite that they build too!