The traditional model for human-robot interaction is at the level of user-interface. For example, given a robot structure (say Robonaut - a relatively complex many DOF human-torso style robot) how should the robot be commanded by the human? We propose that for planetary exploration on the scale of Martian or Lunar habitation, a fixed robot/architecture will not be versatile enough for the variety of known and unexpected events. The astronauts will need to construct the tools to complete tasks from a rich set of modules. There needs to be new programming paradigms that suit the rapid resource constrained contingencies an astronaut may face.