One possibility is to look at the non-canonical amino acid (NCAA) work with Rosetta, and see if you can set up your residue as a non-canonical.
I’m not sure how well that would work, though. Support for non-canonicals in Rosetta is still somewhat in it’s early days, and has focused so far on “peptide like” non-canonicals. You may run into issues with your ligand not being peptide-enough to work with the NCAA framework. (Also, I don’t know how much support pepspec has for NCAAs. It might be okay, but there are often embedded assumptions in certain applications which limit how well NCAAs are handled.)
Another possibility is to model the peptide and the ligand as separate molecules, and then actually make the bond between them as a post-processing step. You can possibly mock the bonding arrangement by putting constraints between the ligand and the protein. Again, how well this might work depends on how the protocol is written.