iStar is a popular goal-based modelling language that has been used in different domains. The language has been extended to incorporate new constructs related to an application domain or to adjust it to practical situations during requirements modelling. However, often iStar extensions are proposed in an ad-hoc way resulting in several problems such as incompleteness, inconsistency and conflicts. This work aims to propose a process to be used by those interested in proposing systematic iStar extensions. A systematic literature review and a qualitative study were conducted previously to identify not only the current iStar extensions, but also to understand what the researchers feel that are important when proposing them. In this paper, we propose a process to support the proposal of new iStar extensions. It is driven by model-based development concepts, reuse of existing iStar extensions and the experts' best practices. It has already been used to propose several iStar extensions. Furthermore, the feedback on the proposal provided by several iStar experts was quite positive. Our process can avoid several problems in current iStar extensions, such as the absence of the constructs meaning, lack of metamodel definition, inconsistencies between abstract and concrete syntaxes and conflicts in the concrete syntax.