I've been planning a system called slopscore for estimating how likely a project is to be vibecoded. It's not an exact science and I'm going for a "best effort" approach, in which we look at a number of heuristics and compute some aggregate score.

Anyways, I'm hoping to write it in OCaml, mainly because I like it but it's also pretty well suited to the task anyways. I plan to model a "repository" as a product type, since then I can write pure functions of repo -> signal and then signal list -> score.

But sadly, OCaml lacks a library for high level Git operations, like "list commits", "diff" types, etc. So I am thinking of writing a separate, small library that wraps the git CLI, parsing its output into record types like commit list. Then I can use this lib in slopscore to clone a repository, "scan" it to construct a repo type, then apply any number of functions to it that compute the signals which I can aggregate.

EDIT: P.S. I would really appreciate some input on these heuristics. Please let me know if you have any recommendations!