An agent can choose to forego bene ts from side opportunities and to instead provide bene ts to the principal. In return, the principal o¤ers rewards. If this exchange is not contractible, typically repeated interaction will be required to sustain it. This model allows the agents productivity in contractible and possibly also non-contractible actions inside the relationship to be correlated with productivity in side activities. This arguably realistic assumption yields several novel implications for the feasibility of relational contracts and for agent selection by principals. The analysis reveals, for example, that optimal agent productivity is often non-monotonic in the importance, to the principal, of ensuring agent reliability.