ticket-triage

ticket-triage categorizes, prioritizes, and routes an incoming support ticket or customer issue, producing a structured triage assessment with a suggested initial response. It takes a ticket text, customer message, or issue description as its argument and is used when a new ticket arrives and needs categorization, a P1-P4 priority, a routing decision, or a duplicate/known-issue check before being handed off.

The workflow runs in steps. First it parses the issue to extract the core problem, symptoms, customer context, urgency signals, and emotional state. Then it categorizes using a taxonomy of nine primary categories — bug, how-to, feature request, billing, account, integration, security, data, and performance — each with signal words, applying tie-breaker rules such as letting root cause drive the category (a login failure caused by a bug is a Bug, not Account) and leaning toward Bug when in doubt. Priority is assigned on a P1-P4 framework: P1 critical (production down, data loss, security breach, most users affected, respond within 1 hour), P2 high (major feature broken, no workaround, respond within 4 hours), P3 medium (workaround available, small impact, respond within 1 business day), and P4 low (cosmetic issues, feature requests, general questions, respond within 2 business days), with escalation triggers that automatically bump priority — SLA breaches, repeated reports, executive involvement, a failing workaround, or expanding scope.

Before routing, it checks available sources — a support platform, knowledge base, and project tracker — for similar tickets, known issues, and existing bug or feature reports, applying duplicate detection. Routing rules map categories and complexity to a queue: Tier 1 frontline for how-to, documented known issues, billing, and password resets; Tier 2 senior support for bugs needing investigation, complex configuration, integration troubleshooting, and account issues; Engineering for confirmed code bugs, infrastructure, and performance degradation; and Product for feature requests. The final output is a structured triage block with category, priority and justification, product area, issue summary, key details (customer, impact, workaround, related tickets, known issue), a routing recommendation, a suggested initial response drawn from auto-response templates, and internal notes. It then offers follow-up actions such as drafting a full customer response, searching for more context, or escalating via a companion customer-escalation flow.