We've published a "Notes on Stake Delegation and UX Security" which defines some of our protocol constraints.

[shielded-nate]

The seed for this document is in #19 (comment) and as we refine that it should lead to a few specific protocol constraints. My guess is this will primarily be around how validators are designated in the protocol (and how that needs to be a self-authenticating identifier which is integrated into a model UX flow that avoids certain kinds of phishing or other secure-UX issues we're aware of).