Getting started
Invite it, grant the right permissions, set a log channel, then watch for a few days before enforcing anything.
Adding Wardnox takes a couple of minutes. Getting it right takes a few days, and almost all of that is waiting rather than working.
1. Invite it
Follow the invite link, choose your server, and approve the permissions.
Wardnox will not say anything when it joins. No setup card, no announcement, no message in whichever channel it happened to be able to post in. A security bot's first act should not be to talk in a room nobody invited it into, so it starts watching quietly and waits for you in the dashboard.
2. Check the permissions it actually got
Four are required, and without them it cannot work at all:
| Permission | Why |
|---|---|
| View Channels | See messages, in order to scan them |
| Read Message History | Read context, and edits to older messages |
| Send Messages | Post the security log |
| Embed Links | Render the log entries |
The rest are optional, and each one buys exactly one action:
| Permission | Enables |
|---|---|
| Manage Messages | Deleting an offending message |
| Timeout Members | Timeouts |
| Kick Members | Kicking |
| Ban Members | Banning |
| Manage Roles | Quarantine |
| Manage Nicknames | Renaming an impersonator |
Grant only the actions you intend to use. Wardnox tells you in the dashboard if a rule is configured to do something it lacks the permission for, rather than failing silently at the moment it matters.
One thing Discord will not tell you: the bot's own role must sit above any member it acts on. A bot with Ban Members still cannot ban somebody whose highest role is above its own. If enforcement seems to do nothing to your moderators, this is almost always why.
3. Set a log channel
Nothing is visible until you do. Detections are recorded either way, but a staff-only channel is where you will actually read them, and reading them is the whole of the next few days.
Make it private. The log quotes what was caught, and a public one hands your members a feed of live scam attempts.
4. Watch, and change nothing
Your server starts in monitor mode. Everything runs — every detector, the full judgement, the log entry — and nothing is enforced. The log records what would have happened.
Leave it there for a few days. Your server has traffic nobody else has, and the only way to know whether the defaults fit it is to watch them not firing on your own members.
While you are watching, turn on logging for events that resulted in no action. It is noisy, and it is the only way to see the near misses.
5. Write the rules only you can write
Wardnox arrives knowing what phishing looks like. It does not know your project's name, your official domain, or which of your channels is the one scammers target.
Start with the domains. Read the log for what your members actually link to, and allowlist those. Then add rules for the impersonation your community specifically attracts. Writing rules covers the one decision that matters most.
6. Enforce
When the log has stopped surprising you, switch enforcement on.
There is no prize for doing this on day one, and the cost of doing it too early is a mass deletion you have to apologise for. Tuning your rules sets out how to tell when you are ready, and what to reach for when something misfires.
What to expect afterwards
Most messages are read and forgotten in the same instant. Nothing is stored for them. You will see log entries only for what the engine found interesting, which on a quiet server can be nothing at all for days.
That is not the bot failing. It is what a quiet server looks like.