> Is high-level behavior the only criteria for diagnosis?

Yes, it is: https://www.autismspeaks.org/autism-diagnostic-criteria-dsm-...

Though, to be more specific, it's even moreso about high-level behavior that has a negative impact, which is highly variable, particularly with environment. I think most would agree that this is extremely messy. But is there a better option?

For practical purposes, it makes some sense to diagnose based on high-level behavior and negative impact on one's life. Describing the relevant what rather than the currently-unknown why.

I do think a more granular approach would have benefits as far as treatment and such goes, though. I know a lot of folks with Asperger's diagnoses were upset when that was rolled into autism. I don't think that was the best distinction to be made, but I think it does show that the people with these conditions often aren't opposed to more granularity (myself included).