I don't think it's derived so much as required. You could write down a non-Lorentz invariant action, but you wouldn't want to, because the resulting theory would lack Lorentz invariance, which is firmly established experimentally.
So if a theory includes action that isn't Lorentz invariant, then it will inevitably have something else in it that's also not Lorentz invariant and is required to be according to the initial hypotheses of special relativity?