The default AM1BCC method is currently recommended by Amber for most purposes.
This appears to be a magic number (for now). Maybe 6 is reasonable for a protein interior?
The default value of 0 assumes the input structure should not be perturbed via minimization. ie, 0 minimization steps results in a single point energy calculation.
The default value of 0.95 was determined empirically by early Osprey developers.