Hi,
my experience with RASPA is not great, but it took me a few weeks to set my system up properly, so I can share with you some advice:
1) in the force_field_mixing_rules.def file you put the non-bonding interactions, which in your case probably means the Lennard-Jones epsilon and sigma values (yes, epsilon/kb as given the SI). Pay attention to the cut-off defined in the simulation.input as well as if there are tail corrections applied to the potential. In force_field.def you explicitly define interactions (meaning you'd have to define C-C, C-O, C-H, etc.). In most cases anyway, the forcefields utilize the mixing rules, so you do not have to do that. In that case, you leave the force_field.def "empty" (not literally empty, there is still some input that says there are no rules to overwrite. Just look at the example forcefields provided with RASPA).
2) in the pseudo_atoms.def file you define (in most cases) the atoms of your adsorbate. That means you define what atoms create your molecule by assigning appropriate chemical elements and masses to the atoms (as you can see in the provided files). The labels used in that file must correspond to the ones used in the molecule.def file later. As you want to study CO2/H2O adsorption, and the paper you linked is about alkanes, you have to find the parameters (both for LJ and charges) somewhere else. It can be TraPPE in the case of CO2, and you have plenty of water models to choose from. The charges should be provided there.
3) charges of the MOFs in the SI: the best option is to include them directly in the CIF file. Look at the CIFs distributed with RASPA, you just define a new field in the CIF and that's it. Of course, make sure you match atoms between your CIF file and the ones in the SI. Your CIF may have more atoms than in the SI: either your CIF does not describe the smallest asymmetric unit or/and your CIF describes different symmetry, hence the difference in the number of atoms. If you have used VASP for the geometry optimization, you can probably get the DDEC charges as well if there is a mismatch between their description of the MOF and yours. You can also use built-in charge equilibration method to get the charges, or use some machine learning-based charges (google for PACMOF, for example).
Again, I may be wrong as I still learn RASPA and MC in general. You will also have to spend a lot of hours learning how to use RASPA.
Cheers,
Mateusz