Protection

Anti-nuke: what it watches, how thresholds work, and how to avoid locking out staff.

Protection watches for destructive actions happening faster than a person plausibly would, and stops the account doing them. It is aimed at a compromised staff account or a malicious bot, not at ordinary moderation.

What it watches

Eleven destructive actions, each armed separately: channel and role deletion, channel and role creation, member kicks and bans, webhook creation and deletion, permission changes, server changes and mass mentions.

Each armed action has a threshold — how many, in how long — and a punishment applied when the threshold is crossed. Attribution comes from the audit log, which is why View Audit Log is required: without it the bot can see that a channel disappeared but not who removed it.

Setting thresholds

Set each threshold above what your staff do on a normal day and below what an attack looks like. Those two numbers are further apart than people expect: a moderator deleting three spam channels in a minute is normal, and thirty in a minute is not.

Punishments

Punishment	Effect	Reversible
Strip roles	Removes every role the account holds, ending its permissions.	Yes — one click from the restore timeline.
Kick	Removes the account from the server.	It can rejoin with an invite.
Ban	Removes and blocks the account.	Yes, by unbanning.
Log only	Records the event and takes no action.	Nothing to reverse.

Start with log only for a week. The record tells you what your thresholds would have caught, which is a far better basis for arming them than a guess.