I think that is a bold claim, is hard to measure how much of the good interactions are such because of the implicit capabilities of punishment by others (social or physical) if people deviates from that. Probably that is embedded deep in the unconcient🤔
I think that people indeed are mostrly nice but not because they are harmless but because we're used to interact with others that can do something to enforce a minimum of nice behaviour in us (and the other way around).
As far as I know there has been reaserch in simulating different strategies in societies to see what set of rules were more stable, "tit for tat" was the fist successful strategy so that should say something about the importance of being able to not be nice some times, but more sophisticated models were finded later, all with some sort of variation of a punishment rule.
I don't fully understand what you thing should be tha basis of the soundness of the protocol should be