I like your solution because it doesn't (necessarily) require a backend and would therefore be much simpler to implement.

There are several libraries that extract article text reasonably well. I believe browser extensions are written mainly in HTML/JS/CSS. Would the following node library suffice? https://github.com/ageitgey/node-unfluff

There's just the question of retrieval speed. I suspect simple HTML retrieval is an order of magnitude faster than normal page rendering due to ads, javascript, etc. So the client-only solution you've proposed might indeed be sufficient!