Within the interface, you could structure the datasets and projects differently:
- have a project "data cleaning" with all datasets you want to clean
- use your celaning recipe in this project only, so it's easy to copy it to use it on various datasets
- expose the cleaned datasets to other projects
Alternatively, a hack would be to create a new recipe in the new project, then edit its json definition (which is data_dir/config/projects/PROJECTKEY/recipes/foo.shaker ) , taking the json definition of your generic recipe as inspiration.
We're thinking about ways to improve this.