Hi,
Some phrases or words are so inappropriate they their use would merit an automatic ban from our community, no review needed.
So, I propose an additional kind of watched word list that results in the automatic ban of the person who posted it.
As an add-on, this could generate a notification for moderators in case there’s a mistake or a need for review.
Out of curiosity, how often do you encounter this situation on your forum?
Hi, great question - totally fair. No, not very often. I received a very hostile email last week, though, and it got me thinking about the issue - if someone wanted to disrupt the community, what would be some additional ways to make that difficult for them to do? And that issue seems to be one that is perhaps applicable to any community manager.
Without speaking about how much the issue occurs, there’s what I think is a fact: Watched Words won’t work against such behavior. If someone wants to post an offensive word, they’ll find a way to do it.
As an admin of a few Discourse forums, I’d handle those manually.
Maybe have a look at Discourse Fingerprint - Browser Fingerprinting Plugin if you think problematic users could be aggressive enough to create multiple accounts to achieve such a purpose.
But now I’m talking more as if it was a support or a community topic… 
Agreed with Canapin
I would just use the silence watched word and if you really wanted to you can use the API. Check all new posts using the /posts.json endpoint and check if it has a word and then suspend with the API.
I didn’t mention Silence which seemed a good option at first because of “First posts of users containing these words will require approval by staff” (emphasize by me).
As for the API, I thought about it (I did a somehow similar thing: Automatically delete users flagged by system?) but didn’t mention it as well since it was really a stretch, coming from a feature topic. 
That said, an auto-ban is always a dangerous feature. I highly doubt such a thing would be officially added to Discourse.
This would be very useful, to beat spammers the only efficient way is to ensure it is easier for the admins/mods to remove their posts/accounts than it is for them to create new accounts/post. Right now, there are minimal bulk ban options available to moderators, bulk as in, banning/deleting multiple accounts easily. There are some IP based bulk options, but it’s not effective against modern spammers as IP is basically not a factor anymore.
With the current watched words, it’s possible to auto-silence a user’s first post if it contains a certain keyword. But if for example, the bad actor has created many accounts and has already done their first post, they can proceed to post and the only option for a moderator is to manually go to every individual post or profile to ban/delete them.
With an auto-ban, we at least could let them walk into our traps and make life easier. Currently, if I add a watched word that auto-flags, blocks or censors to target the second posts onward of a spammer, they can simply change the text they post and their accounts are unscathed.
Real and frequently occurring problem, had someone create over 10k accounts last weekend and it’s impossible to know the words to silence in advance of their first posts and by then it is too late. The only way for a moderator to clean up an ongoing spam attack after the spammer has already made many accounts with first posts is very inefficient.
Auto-ban on certain keywords would help a lot, as one tool that could be used in certain cases (they can circumvent this with several methods, but at least it is a bit harder for them and it covers some cases). It isn’t that dangerous because users can simply be unbanned.
Just my 5c cheers.