Part 2 is a very dumb take.

The objetive of fido2 is to avoid phishing in the scenario of a third party actor trying to authenticate from a terminal that isn't the origin of the token emission, since it has origin binding.

So obviously if the third party actor is acting from a compromised terminal with access to the fido2 key, they can be the origin of the request.

Sources & more info to educate yourself in the matter:

Lawrence systems TLDR version: https://www.youtube.com/watch?v=F_E2LZK-bFk

RSA conference: https://www.youtube.com/watch?v=aMo4ZlWznao

Fido alliance: https://fidoalliance.org/specs/u2f-specs-master/fido-u2f-ove...