Hi Team,
I have 3 critical issues with my Comment-to-DM campaign:

1. AI Leaks System Prompts: Instead of waiting silently for a user's reply, the AI actually sends my internal prompt instruction to the user's Messenger: (No further response needed until user replies "OK"). Please fix this LLM parsing flaw.

2. Webhook Double-Triggering: The system triggers my Custom Function webhook twice per comment—first without the keyword, then with it. This causes API conflicts.

3. Feature Request - Custom Public Replies: Currently, the public comment auto-reply is hardcoded ("We have sent you a DM..."). Due to Meta's strict spam filters, DMs often go unseen in "Message Requests". We urgently need to customize this public reply to tell users to manually DM us if they don't see the automated message.

Thanks!
Danny Zhang