At this point, I'm not even sure how this is better than putting it completely to fiat. The DCs (assuming they end with Spellcraft or Knowledge(casting type) checks are probably too low for anyone to fail except with hilariously complicated spells. The example is also a little odd because the skull, although part of the spell, is only part of it superficially. Sure, I have a fire shooting skull, but I could just have an orb of fire shaped like a skull, and as it is, it seems to be better. Breaking it up into tiny parts, then forcing the experiment after 3 parts can pile into a mess depending on the elements you're adding in, although logic might state to do a bunch of simple steps and test, or to test after an extremely complicated part, normal testing aside.

This, unfortunately, also says nothing about what level the spell should be, for example.