I see no documents in the Dovecot documents wiki on how Dovecot features can be used in the fight against spam. One hit came up in a search, and it was about how to fight spam in the MoinMoin Wiki itself (e.g. TextCHAs and such).
I already asked in another thread about removing message files at the search from a designated "learn-spam" folder. My next question would be how to automatically place detected spam into a specific folder for the user to access or ignore as they deem fit. That would seem to me to be something for the deliver command and its -m option. But this might be more of a Dovecot and MTA coordination issue, too (e.g. how to get the spam detection done in the MTA ... Postfix in my case ... to become a label of some sort that ends deciding the folder. The -m option might be harder to get it dynamically set since it would be hard coded in master.cf for Postfix (so a variable has to be able to direct it to INBOX in this case).
But maybe the "standard" way of doing this is to shim another program between the MTA (Postfix for me) and the LDA (Dovecot deliver) to dynamically add the -m option on the fly? Lacking any other clever solution, this seems to me to be the way I could make it work. Maybe there is a special header deliver could detect, such as "X-Spam: yes"?
A wiki document on how Dovecot helps (or how to help Dovecot) in the fight against spam, I think, would be helpful.