Does power-sharing promote peace? Relying on credible commitment theory, past research has predominantly focused on one aspect of this question. Namely, whether power-sharing prevents the recurrence of battle violence between agreement signatories. However, this disregards a phenomenon that plagues post-war countries across the globe: battle violence perpetrated by armed groups outside of the negotiated settlement against the post-war order. To explain this violence, I argue that we have to focus on how power-sharing redistributes power and access to resources across elites in a post-war country. By determining who gets what, when, and how, power-sharing determines the state’s counter-insurgency capabilities and thus shapes incentives and constrains for extra-agreement battle violence. Personalized power-sharing, for instance, gives elites privileged access to state resources, facilitates effective counter-insurgency strategies, and thus decreases extra-agreement violence. In contrast, structural power-sharing limits elites’ access to resources and their ability to prevent armed challenges resulting in higher levels of violence. To empirically test these propositions. I combine data from the Power-Sharing Event Dataset (PSED) with the UCDP Georeferenced Event Dataset (GED) for peace agreements in Africa and Asia signed between 1989 and 2006. I analyze this data using count models, matching procedures and correlated random effects models. The empirical results support my expectation that personalized power-sharing is associated with fewer extra-agreement battle deaths while structural arrangement facilitates post-war rebellions. This study contributes to an improved understanding of power-sharing as a conflict resolution tool and highlights its divergent effects on actors inside and outside of peace agreements.