Keyword-triggered escalation recipes are prone to false positives the moment a customer uses the trigger word in a low-stakes sentence, and duplicate tickets from one frustrated customer can quickly turn into alert spam. This prompt gets both problems on the table before the recipe goes live.
Automation
Escalate urgent support tickets to Slack in real time
A scenario prompt that converts a described ticket-escalation workflow into a recipe, resolving priority-detection, on-call rotation, and duplicate-escalation edge cases.
Works with Claude / GPT670 uses★ 4.3
The prompt
automation-support-ticket-to-slack-escalation
A support team describes their process like this: "When a customer submits a support ticket tagged 'urgent' or mentions 'refund' or 'cancel' in the message, I want it posted immediately to our #escalations Slack channel and assigned to whoever is on-call, so nothing urgent sits in the queue overnight." AVAILABLE TOOLS: [Help desk ticketing system (trigger), Uncanny Automator, Slack, on-call schedule tool] Turn this into an implementation-ready automation recipe, thinking through what the plain description leaves ambiguous: 1. Trigger: the exact combination of tag-based and keyword-based matching, and how to avoid false positives, since a ticket that says "please cancel my duplicate order, everything else is fine" would trip the keyword match but isn't actually urgent. 2. Conditions: state the exact logic for pulling "whoever is on-call" from the schedule tool, including what happens at the exact boundary when a shift changes mid-ticket, so the ticket doesn't get assigned to someone who just went off duty. 3. Actions in order: post to Slack with the ticket summary and a direct link, then assign the ticket in the help desk tool, then tag the on-call person by their actual Slack handle, not just their name, so the notification pings them. 4. Failure handling: if the on-call schedule tool is unreachable and the current on-call person can't be determined, should the ticket still post to Slack unassigned with an @here, or fall back to a fixed backup contact. 5. One edge case specific to this exact workflow: the same customer submitting three "urgent" tickets in ten minutes about the same issue. Decide whether each one re-escalates and re-pings, or whether the recipe should detect the duplicate and thread it under the first escalation instead. Replace the workflow description and available tools with your own.