As abusive supervision entails negative outcomes for individuals and organizations, a better understanding of leader- and follower-related antecedents of abusive supervision can help organizations prevent destructive leadership. In an experimental vignette study with 140 leaders, we tested an integrative model that includes leaders’ narcissism as an antecedent of their abusive supervision intentions. We also tested for the moderating role of followers’ behavior and indirect effects via leaders’ evaluations of followers. We employed the narcissistic admiration and rivalry concept (NARC) to distinguish between adaptive and maladaptive dimensions of grandiose narcissism and found that the maladaptive dimension, narcissistic rivalry, predicted abusive supervision intentions. This effect was strongest when followers behaved dominantly. Finally, we found preliminary evidence that leaders’ evaluations of followers’ likeability, but not of followers’ competence, mediated the relationship between leaders’ narcissistic rivalry and abusive supervision intentions. These indirect effects were not conditional on followers’ behavior. We discuss these findings in light of theoretical and practical implications for individuals and organizations.