The unfortunate answer is, there isn’t a good way to do it. No journal would accept a paper on just the scoring application by itself, and generally none would accept something just documenting the existing scorefunction either. We as a community are frustrated by the lack of a citation for what you want, too.
Immediately, your best bets to cite either the Rohl methods in enyzmology paper (PMID: 15063647) which discusses most of the scorefunction terms, or the Kuhlman paper on top7 which discusses them in the supplemental (PMID:14631033). There is a paper coming (O’Meara et al) which will describe the scorefunction balance in detail (along with a method for scorefunction optimization, which is what makes it interesting and novel for publication) and will serve wonderfully for this purpose…but it’s not out yet.