I am very new to Rosetta and PyRosetta. But I just had some experience on alanine scanning via PyRosetta.
On the webpage of Pyrosetta (http://pyrosetta.org/scripts.html), you should be able to find the sample PyRosetta code (Alanine-scanning and ddG calculation) done by Sid Chaudhury. It will give you the idea.
I asked the same question about specifying environment-dependent H-bond in the scoring function to the MiniRosetta mailing list. Based on the feedback, it seems not possible now to make the exact scoring function in MiniRosetta or PyRosetta as the terms introduced in the PNAS paper. If you want the exact same H-bond weights in the scoring function, Robetta is the only choice.
However, I was also told that the “environmental weights” were somehow “smoothed” from 0.2 for fully-exposed hydrogen-bonding pairs to 1.0 for fully-buried hydrogen bonding pairs, and will be active if you the “score12” score function. I think Liz Kellogg has already used this in her ddg protocol of MiniRosetta to predict the globular protein stability after mutation. But it seems no benchmarks are available regarding to the interface ddg (binding affinities of alanine scanning).
I hope these information will be helpful.