The cut-off energy depends on the pseudopotentials, so if you're using the same pseudopotentials you should use the same cut-off energy.
The k-points sample the reciprocal-space of your simulation cell. If you make a supercell, the real-space cell is increased in size by some multiple, let's call it X. The corresponding reciprocal-space cell as been *reduced* in size by the same factor X, and since you don't have as much reciprocal space to sample you don't need as many k-points; in fact, if you used to have N_k total k-points you now need only N_k/X (although that may not be an integer, so you have to think carefully about exactly what the appropriate sampling would be).