Estimates of demographic rates for animal populations and individuals have many applications for ecological and conservation research. In many animals, survival is size-dependent, but estimating the form of the size–survival relationship presents challenges. For elusive species with low recapture rates, individuals’ size will be unknown at many points in time. Integrating growth and capture–mark–recapture models in a Bayesian framework empowers researchers to impute missing size data, with uncertainty, and include size as a covariate of survival, capture probability, and presence on-site. If there is no theoretical expectation for the shape of the size–survival relationship, spline functions can allow for fitting flexible, data-driven estimates. We use long-term capture–mark–recapture data from the endangered San Francisco gartersnake (Thamnophis sirtalis tetrataenia) to fit an integrated growth–survival model. Growth models showed that females reach longer asymptotic lengths than males and that the magnitude of sexual size dimorphism differed among populations. The capture probability and availability of San Francisco gartersnakes for capture increased with snout–vent length. The survival rate of female snakes exhibits a nonlinear relationship with snout–vent length (SVL), with survival flat between 300 mm and 550 mm SVL before decreasing for females between 550 mm and 700 mm SVL. For male snakes, survival decreased for adult males >550 mm SVL. The survival rates of the smallest and largest San Francisco gartersnakes were highly uncertain because recapture rates were very low for these sizes. By integrating growth and survival models and using penalized splines, we found support for size-dependent survival in San Francisco gartersnakes. Our results have applications for devising management activities for this endangered subspecies, and our methods could be applied broadly to the study of size-dependent demography among animals.