Right now, test vector processing has to be told what type of test vector it is consuming. We should consider adding a field in a test vector that declares what type of test vector it is. Say something like "type": "tree_math".
It might also be worth producing a unified test vector file format, that pulls together multiple individual test vectors of various types:
{
"tree_math": { ... },
"crypto_basics": { ... },
// ...
}