Agents interactions which are inspired from generic protocols require a special approach for modelling, designing and execution. They raise up issues related to their reusability, configuration and instantiation. Agent-oriented design methodologies do not separate the protocols from the private part of agents architecture. Thus, none of these parts (public and private) are reusable. Furthermore, these methodologies do not provide any guidelines to configure interactions from generic protocols. Hence inconsistencies are hard to anticipate and solve. In this paper, we propose a method to address these issues by automatically deriving agents coordination mechanisms from generic protocols. This method decouples the protocols from the remainder of agents architecture. It also reduces inconsistencies during interactions.