Naive anti-Molochism would be something in the neighborhood of absolute pacifism. Doing things with the highest expected global utility, without paying attention to the iterated meta-level. You might not last very long that way.
Proper anti-Molochism would probably be fairly pacifist, but it would be less naive, and maximize utility over the long term. That might mean building a coordination structure that can outcompete strategies that don't involve joining that structure. But that's difficult, because you have to balance between fitness (Molochian) and utility maximization conditional on success (anti-Molochian).
You need to find a decent Nash equilibrium.