SkullBot
Moderation
Scam image detection: fake giveaways and hijacked accounts
Short answer
Turn on the Scam image detection module in Moderation and describe the scam pictures. The bot checks the first image of a message after the normal filters and removes it when the AI is sure it is a scam.
Step by step
- Check VIP+The module only works on servers with an active VIP+ slot.
- Turn it onIn the dashboard go to Moderation, pick Scam image detection and switch it on.
- Describe the scamThe first description (MrBeast) is ready. Add more descriptions of scam pictures, one per line, up to 10.
- Set threshold and penaltyKeep the confidence threshold at 80 or change it. Pick the penalty for the author: delete only, timeout for a set time (you give the minutes), kick or ban.
Why a second layer
Scammers and hijacked accounts repeat the same images: a photo of a famous person, text about a prize and a QR code or link. Link filters miss this because the whole bait is inside the picture.
How the bot keeps costs low
- It first waits briefly so the normal filters (image spam, links, automod) can remove the message. A removed message is never sent to the AI.
- It analyzes only the first image of a message, even when there are several.
- It compresses the image to 640 pixels and JPEG, and the AI receives it at low detail.
- The same image with the same descriptions is judged once, so a repeated scam does not cost twice.
Penalty and logs
The image is always removed. The author can also get a timeout for the minutes you choose, a kick or a ban. Every detection goes to the log channel (Modlog module) as an embed with the author, channel, penalty, AI confidence, matched description and an image preview, and the penalty is saved to /cases.
AI channels
When an image lands in an AI channel or comes with !ai, it is judged as a possible scam first and only then does the AI answer. A detected scam is removed and the bot does not reply to it at all.
Frequently asked questions
Does it work on every server?
No. The module is for servers with an active VIP+ slot (plain VIP is not enough) because every image check costs an AI request.
What if the AI is wrong?
An image is removed only when the AI confidence passes your threshold. Raise the threshold to remove less, or make the descriptions more precise.
Does the bot check every image in a message?
No, only the first one. That is deliberate to save cost, since scams usually put the bait in the first image.