We added sparkcompute, sparksink, and aggregator plugin types in 3.4. There needs to be an archetype to help people get started creating custom plugins.
The pattern we've been following is that there is an archetype for each plugin type. This may become hard to manage as plugin types increase, and there actually isn't much of a difference between the archetypes. I may just consolidate them into a batch archetype and a realtime archetype.
Added a cdap-data-pipeline-plugins-archetype to help users get started writing custom plugins for the new data pipeline app.