agent-email-inbox

What problem does it solve? AI agents that process inbound email receive untrusted input, and without sender validation, content filtering, and webhook verification, anyone can email instructions that your agent will execute. This Skill provides the security patterns and setup steps to receive and act on email safely. ## Core Features & Use Cases - Webhook-Based Receiving: Set up Resend inbound email with email.received webhook events, signature verification via Svix headers, and raw-body handling in Next.js or Express. - Five Security Levels: Implement strict sender allowlists, domain allowlists, content filtering, sandboxed processing, or human-in-the-loop approval for untrusted email. - Local Development Tunneling: Expose local webhook endpoints with Tailscale Funnel, ngrok, or Cloudflare Tunnel for testing. - Use Case: Build an automated support handler where emails to [email protected] trigger your AI agent, but only messages from allowlisted senders are processed while all rejections are logged and reported to the owner. ## Quick Start Ask the AI to set up a secure Resend email inbox webhook with a strict sender allowlist so your agent can receive and respond to emails.