I considered that, but it feels like a hack (and it’s just a prototype, so I can afford some complexity). There’s no new key type or even key encoding introduced, and any future new key type should include the same replay-protection.
Also, IIUC, that approach won’t work with key-path spends.