Auto-Mod

Delete spam, scam links, invites, slurs, and pictures you have blocked.

What it does

Auto-Mod watches every message and acts the instant one breaks a rule you set — deleting it, warning the member, muting or kicking them, and logging what happened. It runs entirely on your configuration; Cyndrix ships nothing enabled by default.

There are thirteen independent modules. Turn on only what your server needs:

• Banned Words — a word list with three matching modes. Smart (the default) matches whole words and catches leet-speak (b4dw0rd), without flagging "classic" for containing a rude substring. Whole is exact-word only; substring matches anywhere.
• Link Filter — block every link, or allow only domains on your list.
• Invite Filter — remove Discord invites to other servers.
• Mass Mentions — trip when one message pings more than N people.
• Anti-Spam — trip on N messages inside a time window.
• Cross-Channel Spam — the scam-blast catcher: the same text, link or image posted across several channels in seconds. It can also require the same content from several different accounts before acting, which is what a raid actually looks like.
• New Member Gate — restrict links, images and invites for accounts younger than your threshold, or members who joined minutes ago.
• Excessive Caps — messages that are mostly CAPITALS. Short ones like "OK LOL" are ignored, and mentions, links and emoji don't count.
• Emoji Spam — more than N emoji in one message (custom and normal both count; a family emoji counts once).
• Glitch Text — z̷a̷l̷g̷o̷-style text buried under stacked accent marks. Normal accents in French, Hindi, Vietnamese and so on are never flagged.
• Repeated Messages — the same message sent again and again in a short window, in any channel.
• File Flood — too many images, files or stickers in a short window.
• Image Scan — upload pictures that should not be posted. A matching attachment is removed and counts once on the strike ladder. Members see “blocked image”.

Image Scan compares PNG, JPEG, WebP, and GIF attachments to the pictures you uploaded. It does not read links, stickers, or videos, and it is not an NSFW detector. A GIF is compared from its first frame. The check runs only when Auto-Mod and Image Scan are both on and the library has a picture. Bots, webhooks, admins, ignored channels, and ignored roles are skipped. NSFW channels are skipped unless you allow them. Files over 8 MB are skipped, and only the first 10 images on a message are considered. Each picture checked costs 1 Action Point, match or not. If the balance cannot cover every image on that message, the message stays up and nothing is spent. On Free this shares the 500-point translation balance. Premium, Ultimate, Nexus, and Professional take it from the server's Action Point pool, including when Nexus or Professional is using its own AI key. The library holds 10 pictures on Free, 20 on Premium, and 100 on Ultimate, Nexus, and Professional. Turn it on in Dashboard → Auto Mod → Image Scan.

The Test a message box on the dashboard shows which modules a message would trip, before you switch anything on.

The strike ladder

Instead of every offence carrying the same punishment, strikes escalate. Strike 1 uses step 1, strike 2 uses step 2, and anything past the last step repeats it. Strikes decay after a window you choose, so an old mistake doesn't haunt a member forever.

The default ladder is delete → warn → mute 10m → mute 1h → kick, but every step is yours to change from Auto-Mod → Defaults → ⚙ Edit ladder: pick the action per step, set mute durations, reorder them, add up to ten, or cut it down to two. Each module chooses individually whether it feeds the ladder — cross-channel scams usually shouldn't, since you want the mute immediately.

Setup

1. Open the dashboard → Auto Moderation, and turn on the master switch.
2. Set a default log channel. Without one, Auto-Mod still acts but silently — you'll have no record of what it removed.
3. Enable the modules you want and open ⚙ Settings on each for its thresholds and word lists.
4. Add bypass roles (Defaults → ignored roles) so moderators aren't caught by your own filters, and per-module ignored channels for places where the rules shouldn't apply.

What members see

Commands

This module has no commands — it's configured entirely from the dashboard.

Related