Skip to main content

The Out-of-Office Disaster

When an AI agent sends an email to a prospect or customer who has an automated responder enabled (e.g. “I am out of the office until Monday”), the auto-reply lands in the agent’s inbox. Without loop detection:
  1. The agent receives the auto-reply.
  2. The agent interprets the email and sends a polite response (“Thanks for letting me know!”).
  3. The recipient’s mail server fires another auto-reply.
  4. The loop repeats indefinitely, burning hundreds of dollars in OpenAI/Claude tokens in minutes.

Detection & Suppression

gork.email automatically inspects incoming RFC headers:
  • Auto-Submitted: auto-replied
  • X-Autoreply: yes
  • Precedence: auto_reply / bulk
  • Repeated identical subject lines within a rolling 10-minute window.
When detected, the webhook payload sets:
If loop suppression is active, outgoing replies back to the looping sender are automatically paused, containing token burn immediately.