I used to do CG work ALL the time in lammps. I don't have example scripts
on me (on my old computer I do). However, it matters what your CG potential
is on how you run it.
Atom_style I have as either full or molecular (if beads are uncharged).
That's because I'd have bonds, angles, dihedrals, and impropers.
So it REALLY matters on what you want to model.
As for charges, put charges in like you would for a normal input file. Go
through the example files and search for the one with a protein. That's a
good one to model it after. If you have neutral beads, make their charge =
For MOST CG potentials, you would use a table file for your non-bonded
interactions. For some (e.g. Martini) just use pair_style lj/cut/coul/long.
If you send more info, I can give you more info, but kind of shooting in the
dark right now on how to help.