It's good, but I'm snagging on the examples here; for instance, the insistence on consistent use of camel or snake case. That seems to me to be purely a tooling limitation; a more code-aware grep could just automatically synthesize the covering regexp for both. It feels like there's a sweet spot in between LSP-ism and grep-ism that isn't being captured.