Performing simulation and data analysis on supercomputers requires using workflow management systems (WMSs). Due to the high diversity of application requirements, there are currently many different WMSs in use with very different input languages and enactment mechanisms. This hinders the reuse of workflows across these WMSs and makes the transition from one WMS to another difficult. We propose a lightweight framework based on a generative approach to make this affordable for the WMS end user. With a proof-of-concept implementation, wfGenes, workflows for different workflows systems are generated automatically from an abstract workflow description. We tested wfGenes for a use case in computational nanoscience to generate workflows for FireWorks and SimStack. The framework can be readily extended to support further WMSs and to adopt a more standardized workflow description language for its input.