In a model of dual-agency problems where borrower–lender and bank–nonbank incentives may conflict, we predict a hockey stick relation between bank skin in the game and covenant tightness. As bank participation declines, covenant tightness increases until reaching a low threshold, at which point the relation sharply reverses and covenant protection is removed with a commensurate increase in spread. We find support for the hockey stick relation with bank’s stake in covenant-lite loans averaging 8% (0% median). We also find that covenant-lite loans are more likely when borrower moral hazard is less severe and when bank relationship rents are high.