Hi Mathias,
It should already be possible to do this by combining cospherical and coplanar (since if the points share both a plane and a sphere they must share a circle).
If this doesn’t work well, a direct co-circular goal would be fairly simple - just like this:
I’d forgotten that there isn’t a direct circle fit method in RhinoCommon.
If using a combination of sphere fit/plane fit/intersect like that, then I don’t think there is actually any advantage to having it in a custom goal - it is probably faster to use the existing co-spherical + co-planar goals, as that way you avoid having to explicitly calculate the intersection curve.
What error are you getting in Rhino6?
It might just be about making sure that the scripted goals are referencing the same copy of KangarooSolver that is being loaded by 6.