I should add that I let traits suggest skill sets so that nothing would be the odd man out. The fussiest part was dividing points in attributes over relatively large numbers of skill sets. Well, that and ascribing motive to a few obscure traits. For instance, some appearance-related perks could be Charm or Deception, depending on whether the player uses them to let a sneaky PC pose as someone else or a charming one to attract people who like a certain class of looks. But I hasten to add that the latter class of things actually didn't change scores by more than 0.1 either way.