n8n Webhook Email Reply Drafting with Groq & Guardrails

Draft safe, compliant email replies from inbound messages—automatically

This n8n workflow receives an inbound message via a webhook, strips hidden instructions and sensitive data, and then drafts a reply using Groq—only from facts you approve. Every draft is validated in code before sending, and the recipient is set from the original sender.

What this workflow does

• Normalizes the inbound webhook payload into consistent fields (sender, subject, body, and message ID).
• Reconstructs the visible text a human would read, detects hidden/encoded instruction-like content, and assigns a risk score.
• Uses n8n Guardrails to replace sensitive values (e.g., card numbers, IBANs, national ID numbers, and API keys) with placeholders so the LLM never sees them.
• Blocks high-risk messages from reaching the LLM and marks them for review with the detected reasons.
• Drafts the reply with Groq as JSON, where the model focuses on wording only and never selects the recipient.
• Verifies the draft in code for length, unapproved links, stray email addresses, and card/key-shaped text—then sends to your configured sending endpoint addressed to the original sender; anything else is held with a reason, and one report covers the run.

Use cases

• Automatically respond to customer inquiries while preventing prompt injection and sensitive data leakage.
• Route risky inbound messages to review before any AI drafting occurs.
• Enforce outbound safety rules (length, links, and data patterns) for SaaS support email replies.

Technical details

• Nodes/logic: webhook, if, set, code, merge, sticky note.
• LLM provider: Groq via a “POST Writing to LLM” node (free tier is sufficient).
• Safety layer: n8n Guardrails for redaction + code-based validation before sending.

Setup: POST your inbound source to the workflow webhook URL and ensure payload field mappings match (or update Settings). Add a Groq API credential to the LLM node.