After creating a pipeline from a Kafka connection, the output schema of the Kafka source is set to a single field named "message". Whereas the Wrangler plugin expects a field named "body".  The consequence is that all data reaches the sink as null (or the pipeline fails if any plugin expects non-nullable input).

Either the Kafka source should emit the "body" field, or the Dataprep should start with a field named "message" after opening the connection.