And if the bank implements the best form of Passkey (and implements it well) in this case it would require the user to biometrically authenticate before sending money (perhaps based on dollar amount or transaction volume) to a new account. The user would hopefully learn over time that the biometric requirement is aligned with money transfers and would grow suspicious if they were prompted for it during a scam session.

I'm not sure I follow this point? A trojan would fail to exfiltrate passkeys (assuming they're stored securely). It would be able to see a valid login and potentially use that session or exfiltrate the session, but wouldn't be able to get persistent account access. This seems better than stealing a password and getting persistent account access, or possibly stealing a TOTP code for the same. It's bad, but I think it's made better by passkeys? Unless I'm misunderstanding your point.