Did you see the reply above from huxz? She’s the first author of the paper. If you want to do it as she did, then you should reply to her above (I’ll let her know to communicate with you further).
If you want to do this in Rosetta3, then be prepared to start writing a fair amount of C++, because (as she said) this protocol wasn’t ported from ++ to Rosetta3.
A) First you’ll need to figure out how you translate your Rose library into Rosetta fragments using the fragments machinery. I don’t know how to do this. I can contact the fragments person to see if he can help if you go this route. This is likely to be the hard part.
Next you’ll need to compose your protocol out of the existing movers (I assume you looked at the Mover part of the Rosetta3 reference). The lower-level movers will work for your purpose (fragment insertion movers, CCD/KIC movers) but the higher-level movers (like loop modeling overall) may not. You might be able to combine your fragment library after step A) with the existing fragments+CCD loop protocol. If you want to try a fragments+KIC protocol (nobody else has written one, because KIC sampling ignores the input conformation unless you tweak the settings) then you’ll need to write one, modeled after the fragments+CCD protocol.
C) If you are writing your own code, there’s no need to separate design from backbone generation, you can just throw design steps into the middle of loop modeling.
I can offer advice and input on and C) but I can’t write the protocol.