The long and short of it is: a bot should NEVER initiate the whisper/conversation.
It lacks “user intent” and is easily mis interpreted as Spam by the filters.
The bot should reply to a command in the same medium that the command was run.
IE:
- user runs command in chat, bot replies via chat
- user runs command in whisper, bot replies via whisper
It’s not an RFC’ed rule, but it’s a very very good guideline, and will solve a LOT of headache.