I don’t think brms's approach is completely documented anywhere beyond the code of the package and has in fact evolved over time. Some relevant discussion and pointers at: How does brms use the given data to adjust the default priors?
I think the philosohpy of brms is that the moment you start to be curious about priors, you should be defining your own anyway. The defaults are not some kind of “best practce”, they are just meant to avoid the model breaking badly for most use cases - nothing more, nothing less.
Best of luck with your model!