The smart check asks whether a message bullies or personally attacks someone. Friendly banter scores lower when the recent messages show both sides are in on it.
Typical catches
- Direct insults, and pile-ons aimed at one member.
- Putting someone down again after they asked you to stop.
- Personal attacks dressed up as jokes, when the context shows only one side is laughing.
Outcomes
Above your action threshold, the message is removed and strike points go on. Between your review and action thresholds, mods get a case card and the message stays up until someone acts.
Tuning
Rather handle this by hand? Switch the category off in /settings. Banter in your server runs hot? Raise the thresholds. Either way, the child-safety and wellbeing checks keep running on their own.