Rough definitions: good for agent X - net positive utility for agent X, evil for agent X - net negative utility for agent X. Or possibly: evil for agent X - a utility function that conflicts with that of agent X.
Good and evil don't have to be "written into the structure of the universe" to be coherent concepts. I assume you make choices. What is your criterion for choice? I also assume that you aren't completely selfish. You care about the welfare of other people at least to some degree right?
Of course, if two people/agents truly have differing utility functions, what is good to one may be evil to the other, but that doesn't invalidate the concepts of good and evil.
That's not 'good and evil', just 'desired and undesired' - much milder and broader concepts.
Tyler Cowen argues in a TED talk (~15 min) that stories pervade our mental lives. He thinks they are a major source of cognitive biases and, on the margin, we should be more suspicious of them - especially simple stories. Here's an interesting quote about the meta-level: