Skip to main content
Antinuke watches the audit log. When someone performs too many destructive actions inside a time window, they get punished and — where possible — the damage gets reverted. It’s aimed squarely at the threat other systems can’t touch: people who already have permissions. A compromised admin account, a staff member who’s decided to burn the place down, a bot with more scope than it should have.

Who can configure it

Antinuke is deliberately locked tighter than everything else. Only the server owner and members explicitly added as managers can change its settings — Administrator alone isn’t enough, because an attacker who gains Administrator would otherwise just turn it off.
Aliases for trust: manager, mod, admin.

Modules

Each module is toggled independently and has its own threshold and window.
Read a threshold as “more than N of this, within this window, gets you punished.”

Modules with no threshold

Any bot added to the server is immediately kicked, and whoever added it is punished. There’s no threshold — one is enough.This is the single highest-value toggle in the whole system. Adding a malicious bot is by far the most common way a server actually gets nuked, because it takes one click and needs no permissions of the attacker’s own beyond Manage Server.Legitimate additions still work — whitelist yourself, then add the bot.
Watches for dangerous permissions being granted to a role — Administrator, Ban Members, Manage Roles, Manage Guild and similar. When one appears, the grant is instantly reverted and whoever made it is punished.This catches the subtler attack: rather than deleting anything, the attacker quietly gives @everyone Administrator and comes back later. Aliases: permission, perms.
Detects mass-mention spam being pushed through a webhook, deletes the messages and removes the webhook. Webhooks bypass most other protections because they aren’t a member, which is exactly why this exists separately.Aliases: wspam, hookspam, whspam, mentionspam, spam.

Punishment

stripstaff is worth considering for a large staff team — it neutralises the account without the drama of a ban, and !stripstaff restore @them puts every role back once you’ve worked out what happened. jail needs the jail role set up first, via !setup. Aliases for the command: action.

Whitelisting

Two different lists, for two different things:
Aliases for whitelist: exempt, wl. Whitelist sparingly. Every entry is a hole in the system, and the whole point is that a compromised admin account isn’t trusted. Bots that legitimately mass-manage things — a ticket bot creating channels, a role bot doing bulk assignment — do need whitelisting or they’ll trip thresholds during normal operation.

Restoring after an attack

When a module fires, the alert includes a restore button. Antinuke snapshots roles and channels before they’re deleted, so restoring recreates them with:
  • The original name, colour, position and permissions
  • Every member who had the role, re-assigned
  • Channel category, topic and permission overwrites
Snapshots are kept for one hour. After that the button stops working and you’re back to backups.

Reviewing the configuration

Aliases: config, cfg, configuration, overview, view, ov. Shows every module, its threshold and window, the punishment, the manager list and the whitelist in one embed.

Command reference

Group aliases: antiwizz, an, aw.