mongoexport --help
....
-f [ --fields ] arg comma separated list of field names e.g. -f name,age
--fieldFile arg file with fields names - 1 per line
You have to manually specify it and if you think about it, it makes perfect sense. MongoDB is schemaless; CSV, on the other hand, has a fixed layout for columns. Without knowing what fields are used in different documents it's impossible to output the CSV dump.
If you have a fixed schema perhaps you could retrieve one document, harvest the field names from it with a script and pass it to mongoexport.