Additional requirements :
- PHP 7.2+, and an API key for any OpenAI-compatible endpoint.

Stops the classic overnight spammer: a fresh account posts a few harmless messages to pass XenForo's spam checks, then floods the forum with link spam.

When a recently registered account posts a message with a link, the add-on sends it to an AI model for a spam verdict:

• Spam -> the message goes to the approval queue, with the AI's reasoning visible to moderators.
• Clean -> it posts normally. Legitimate new users sharing a link aren't punished.
• API down -> fail-closed: the message is held anyway, so an outage never opens the door to spam.

Extra protections:

• Queue flooding: once a recent account has several messages in the moderation queue, further submissions are blocked outright - no AI call needed.
• Posting bursts: link-free bursts of short messages from a watched account are also analysed.
• Media embeds: hidden links in [MEDIA] embeds are held for review (trusted platforms like YouTube or Imgur can be whitelisted).
• Closes the message-count loophole: spam checks stay active for a number of days from the user's first message, not from registration - sleeper accounts don't slip through.
• Prompt-injection resistant: the message is treated as text to classify, not instructions to follow.

Works with any OpenAI-compatible API (OpenCode GO, DeepSeek, OpenRouter, OpenAI...). The AI is only called for recent accounts posting links, so cost is effectively nearly zero and normal traffic gets zero extra queries, no template modifications and no JavaScript.

Everything is handled through XenForo's native tools: spam trigger log, approval queue, spam cleaner and moderator log. All thresholds are configurable under Options -> Limpia Cucarachas.

Télécharger V1.0.4 :