We study a repeated principal-agent model with transferable utility, where the principal's evaluation of the agent's performance is subjective. Our focus is on equilibria which are robust to the addition of small privately observed shocks to the payoffs. Existing constructions of positive-effort...