My new dovecot covers spam/ham training through the standard Spam_train and Notspam_train mailboxes. When a message is copied (I’m using macOS/iOS standard Apple clients) it triggers the spam/ham learning by rspamd.

So far so good.

When I move a message to Spam_train I’d like it to be removed immediately after learning.

When I move a message to Notspam_train, I’d like it to move back where it came from after having been piped

The first is I think doable, but how? Is the second doable?