About this prompt
n8n Workflow From a Plain-English Spec
If you can describe an automation as "when X happens, do Y", you can build it in n8n — everything else is decoration on that grammar. The part people get wrong is the spec: vague triggers, steps out of order, no idea what "worked" looks like, and no plan for what happens when it fails at midnight. This prompt is a five-line spec that maps onto n8n's model and returns the node list with the upstream field each node reads, importable JSON, an error path and a test plan.
How to use. Write the trigger as an event in a named system, the steps in order, the branching conditions, the outcome you will check, and the systems involved. Credentials are referenced by name and never pasted. Paste the output into n8n via Import from clipboard, create the named credentials, run the test plan once with sample data, then activate.
What to change. For scheduled jobs, the trigger is "Schedule Trigger, 09:00 daily" and the first node is a fetch. For AI steps, add "an AI Agent node with tools: ..." to the steps and name the stop condition. If the full version exceeds eight nodes, build the v1 the prompt recommends first.
Not a fit. Processes you have not run by hand yet — manual exposes the edge cases, automation amplifies them.

