Quality diversity algorithms are evolutionary algorithms that aim to evolve diverse repertoires of high-quality solutions. Quality diversity has recently been used with considerable success to evolve repertoires of single-robot controllers in a wide range of applications. In this paper, we propose a methodology for the evolution of repertoires of general swarm behaviours. We use a quality diversity algorithm that relies on a behaviour characterisation and a quality metric that are task-agnostic, meaning that the repertoire evolution is not driven towards solving any specific task. We use a total of eight swarm robotics tasks to evaluate the behaviours contained in the evolved repertoires a-posteriori, and compare their performance with direct task-specific evolution. We show that the repertoires contain a wide diversity of swarm behaviours, and for most of the tasks, the behaviours in the repertoire have a performance close to the performance achieved by task-specific evolution.