Can religiosity sway a society's propensity for violence against outgroups? We first introduce two state-year-level religiosity measures for several pre-Enlightenment European states with the frequencies of (i) religious language in book publications and (ii) Christian names of newborns. To identify causal effects on warfare, we exploit the local visibility of solar eclipses – phenomena orthogonal to climatic, cultural, economic, environmental, and institutional developments that, in pre-Enlightenment Europe, were overwhelmingly viewed as supernatural, religious events. Accounting for dyad- and year-fixed effects, we observe positive, statistically significant, and quantitatively sizeable effects on subsequent attack war onset. Reduced form estimates, robustness checks (e.g., acknowledging dyad-specific time trends), and placebo exercises yield consistent patterns. Exploring mechanisms, religious terminology explicit to religious outgroups (specifically Jews and Muslims) spikes in solar eclipse years and predicts attack war onset, particularly against Islamic states. Finally, consistent with the idea of a religious primer highlighting ingroup-outgroup demarcations and exacerbating tensions along such lines, city-year-level solar eclipses also predict (i) Jewish expulsions and (ii) witch trials in pre-Enlightenment Europe.