update: sklearn trainer train trigger
The Sklearn Train box gets the train trigger from receiving a OVStimulationEnd or OVStreamedMatrixEnd chunk. The acquisition client does not produce such chunk when the scenario stops. The generic or gdf stream writers don't add a END chunk on stop. The generic and gdf stream readers don't add a END chunk when the data has been fully read. Only the CSV Writer will produce such chunk at the moment.
Therefore, OpenViBE is not ready to support the use of END chunks to trigger actions.
The easier step would be to change the trigger mechanism of the sklearn trainer box to work like the Classifier trainer box for example.