9 Checks Before an n8n LLM Node Sends Anything to a Customer. I Replayed 182 Sends Through Them: 97 of 103 Bad Ones Held, 6 Slipped Through
Last time I shared a 12-point checklist for switching an n8n LLM workflow on. This one is narrower and stricter: what has to be true before an LLM node is allowed to send something to a customer, whether that is an email reply, a public review reply, a refund, or a page to the on-call engineer. I didn't want to just write the list, so I tested it on data I already had. Two of my earlier tests…
Before an n8n LLM node can send anything to a customer, nine strict checks must be met. These checks were tested on previous data with a 9-step list. The tests involved 182 potential sends, with 97 out of 103 held and 6 slipping through. The checks ensure no money-moving actions, refunds or credits going to a human queue, no paging, no internal references or phone numbers, link allowlist, no promises of money or codes, the reply referencing the original message, length cap, and a kill switch with a send log.
The six bad sends in the 182 were mainly refunds without a request, on-call pages on non-urgent emails, promises of refunds, foreign links, and internal secrets or phone numbers. The two checks for bot phrases and specific replies did not hold as intended, but could be a human review score or limited to specific phrases. The main gate that blocks harmful sends is a code node that checks for allowed links, money promises, secret references, length, and a kill switch.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.