rexxzyid/Elaina-AnchorGuard ? reverse-engineered prompt

Build me a small JavaScript package for WhatsApp bots that helps stop crash messages before they can force close the app.

I want it to work with a Baileys socket, watch incoming messages and also check outgoing ones, then automatically delete anything that looks dangerous. If the message came from me, it should revoke it, and if it came from someone else, it should delete for me. It should also be able to block the sender if I turn that on.

Please include a simple way to create the guard, a way to stop it later, and a manual function to check a message payload by itself. Make it catch the common bad patterns like weird invisible characters, huge text, too many mentions, broken button or list payloads, invalid location values, deep nesting, and other malformed message structures.

Keep it best effort, not perfect, and make it easy to customize thresholds and hook into a callback when something is detected. If you need current Baileys docs, look them up online.