The problem of designing workforce shifts and break patterns is a relevant employee scheduling problem that arises in many contexts, especially in service industries. The issue is to find a minimum number of shifts, the number of workers assigned to them, and a suitable number of breaks so that the deviation from predetermined workforce requirements is minimized. We tackle this problem by means of a hybrid strategy in the spirit of Large Neighborhood Search, which exploits a set of Local Search operators that resort to a Constraint Programming model for assigning breaks. We test this strategy on a set of random and real life instances employed in the literature.