Are retraining aliases safe?

[Phil Stracchino]

I switched to rspamd from Dspam last year sometime.  Dspam worked great, but with no active developers the bitrot was getting worse and worse.  However, it had a very simple mechanism for spam/ham retraining — you could just go to the history log in the web GUI, select an incorrectly classified message, and tell Dspam to retrain it as ham or spam.  (But among the bitrot was the problem that redelivery from the spam quarantine for messages retrained as ham no longer worked due to UTF8-related issues.)

So anyway, I recently started looking up how to retrain misclassified messages with rspamd, and I see it can be done from the command line with rspamc.  However, not every user in my household has rspamd installed locally or has a login on the server that runs rspamd.  None of the kids even runs Linux.  (Yet.)

Is it "safe" to set up retrain-ham and retrain-spam aliases in my mail system that feed a mail message to rspamc and have people redirect false negatives to those aliases?  Or will this have undesirable effects upon rspamd's spam detection?

For instance:
retrain-spam:  |/usr/bin/rspamc learn_ham

Is there a how-to document anywhere that covers how to make the most effective use of rspamd, including retraining of false positives and negatives?