@crandall I can't speak to the underlying software scanner, but there's pretty clearly a lot of of subtlety. For example, a repo with no claude commits but CLAUDE.md in .gitignore seems to imply that the author is using claude but not wanting to reflect that in the commits or repo - warranting a lower score than BoxBox (where about 3% of commits have some attribution to claude).

https://slopscan.ava.pet/repo/https%3A%2F%2Fgithub%2Ecom%2Fspacejam%2Fsled/main