I'd use the Schema feature from a FeatureReader, exploding the attribute list and then using name (attribute name) and native_data_type to build a text file? This could be done when reading, or add an extra FeatureReader (Schema feature only) after writing out the data.