mharisbaig/ai-safe-business-automation — reverse-engineered prompt

Build me a secure AI business automation platform for small businesses in Pakistan, especially for customer support teams using WhatsApp.

I want users to be able to log in with different roles like admin, support agent, and viewer. Customers should be able to send complaints through WhatsApp, the system should turn them into helpdesk tickets, classify them, and suggest AI replies. Before anything risky goes out, it should go into a human review queue where a support person can approve, reject, or edit the response.

Please include safety guardrails for prompt injection, private data leaks like CNICs, phone numbers and emails, harmful content, and unreliable AI answers. I also want an admin dashboard with tickets, review queue, audit logs, analytics, and system status.

Make it runnable locally with environment variables, MongoDB, optional Redis, tests, and Docker setup. Use the existing JavaScript style from the repo, and look up current docs online if you need to.