ticket-triage
Classify, prioritize, and route incoming support tickets by extracting intent and entities, assigning severity, and generating initial responses. Use when the user requests ticket triage or provides relevant inputs for this workflow.
git clone --depth 1 https://github.com/seb1n/awesome-ai-agent-skills /tmp/ticket-triage && cp -r /tmp/ticket-triage/customer-success/ticket-triage ~/.claude/skills/ticket-triageSKILL.md
# Ticket Triage Automatically classify, prioritize, and route incoming customer support tickets to the right team with a suggested first response. This skill processes raw ticket text, identifies the customer's intent and key entities (product area, account tier, error codes), assigns a category and priority level, then routes to the appropriate team while drafting an empathetic initial reply. ## Workflow 1. **Receive and parse the ticket** — Ingest the raw ticket including subject, body, customer metadata (account tier, plan, tenure), and any attachments or screenshots. Normalize the text by stripping signatures, quoted replies, and boilerplate. Extract structured fields like order IDs, error codes, and URLs. 2. **Extract intent and entities** — Analyze the cleaned ticket text to determine the customer's primary intent (reporting a bug, asking a question, requesting a feature, disputing a charge). Identify named entities such as product names, feature areas, API endpoints, and specific error messages. Tag sentiment as positive, neutral, negative, or urgent. 3. **Classify into category** — Map the extracted intent to one of the standard categories: `billing` (charges, invoices, refunds, plan changes), `bug` (errors, crashes, unexpected behavior), `feature-request` (new functionality, integrations, enhancements), or `how-to` (setup, configuration, usage questions). If a ticket spans multiple categories, assign the primary and note secondaries. 4. **Assign priority using impact and urgency** — Score priority from P0 to P3 using a matrix. P0: production outage or data loss affecting multiple customers. P1: major functionality broken for a single enterprise account. P2: degraded experience with a workaround available. P3: general questions or minor cosmetic issues. Factor in account tier — enterprise accounts get a one-level priority bump. 5. **Route to the appropriate team** — Map the category and priority to a team: billing tickets go to the Finance Support queue, bugs route to Engineering Triage (P0/P1) or Product Support (P2/P3), feature requests go to Product Management, and how-to tickets go to Customer Education. Attach relevant context and internal notes. 6. **Generate initial response** — Draft a first reply that acknowledges the issue, sets expectations on response time based on priority, includes any immediate self-serve resources (KB articles, status page links), and asks clarifying questions if the ticket is ambiguous. Match tone to sentiment — empathetic for frustrated customers, concise for technical users. ## Usage Provide a support ticket (or batch of tickets) and ask the agent to triage them. Include customer metadata when available for better priority assignment. ``` Triage these support tickets and output a classification table with category, priority, assigned team, and a draft first response for each. ``` ## Examples ### Example 1: Batch triage of 5 tickets **Input:** 5 support tickets from the past hour. **Output:** | # | Subject | Category | Priority | Team | Suggested Response | |---|---------|----------|----------|------|--------------------| | 1 | "Can't login since this morning" | bug | P1 | Engineering Triage | "Hi Sarah, I'm sorry you're locked out. I've escalated this to our engineering team and they're investigating now. As a workaround, try clearing your browser cache or using an incognito window. We'll update you within 2 hours." | | 2 | "How do I export data to CSV?" | how-to | P3 | Customer Education | "Hi James, great question! You can export to CSV from Settings → Data → Export. Here's our step-by-step guide: [link]. Let me know if you hit any snags." | | 3 | "Charged twice for March" | billing | P2 | Finance Support | "Hi Priya, I apologize for the duplicate charge. I've flagged your account for review and our billing team will process the refund within 3-5 business days. You'll receive a confirmation email." | | 4 | "Would love Slack integration" | feature-request | P3 | Product Management | "Hi Tom, thanks for the suggestion! A Slack integration is something we're hearing a lot about. I've added your vote to our feature tracker and will notify you if it moves to our roadmap." | | 5 | "Dashboard is down for our entire org" | bug | P0 | Engineering Triage | "Hi Alex, this is our top priority right now. Our on-call engineering team has been paged and is actively investigating. Check status.example.com for live updates. I'll follow up within 30 minutes with a status." | ### Example 2: Urgent production outage escalation **Input:** "URGENT: All API requests returning 500 errors since 14:32 UTC. Affecting our production environment. 50,000+ end users impacted. Enterprise plan." **Output:** - **Category:** bug - **Priority:** P0 (production outage, multiple end users, enterprise account) - **Team:** Engineering Triage → Immediate page to on-call SRE - **Escalation path:** Auto-notify VP Engineering, assigned CSM, and account executive - **SLA clock:** 15-minute first response, 1-hour status update cadence - **Draft response:** "Hi — this is marked as our highest priority. Our SRE team has been paged and is investigating the 500 errors starting at 14:32 UTC. We'll provide our first status update within 15 minutes. For real-time updates, monitor status.example.com. I'm your point of contact until this is resolved." ## Best Practices - Always extract customer account tier before assigning priority — enterprise customers with production impact should never be lower than P1. - Use the secondary category tag for tickets that span billing and bug (e.g., "I was charged for a feature that doesn't work") to ensure both teams have visibility. - Pre-populate the initial response with relevant KB article links by matching extracted entities against your knowledge base index. - Re-triage tickets that receive a customer reply within 1 hour, as additional context often changes the category or priority. - Log triage decisions with reasoning so the mode
Design reproducible evaluations for AI agents with representative task sets, explicit rubrics, appropriate graders, baselines, regression gates, and failure analysis. Use when defining agent quality, comparing prompts or models, validating a release, measuring tool-use reliability, investigating regressions, or deciding whether an agent is ready for production.
Design privacy-aware observability for AI agents using traces, spans, structured events, metrics, cost attribution, dashboards, alerts, and investigation workflows. Use when instrumenting an agent, debugging intermittent tool or model failures, defining service-level objectives, analyzing latency or spend, auditing agent decisions, or preparing production monitoring.
Design and verify auditable human oversight, approval gates, escalation paths, and safe state transitions for AI agent workflows. Use when deciding which agent actions require review, adding approve/reject or dual-control flows, preventing unauthorized autonomous effects, creating decision records, reducing rubber-stamping, or recovering safely from rejected, expired, or failed actions.
Design, implement, harden, and verify Model Context Protocol (MCP) servers with precise tool contracts, least-privilege authorization, safe transports, structured errors, and interoperability tests. Use when creating a new MCP server, exposing an API or data source through MCP, reviewing an MCP server design, adding or revising MCP tools, or preparing an MCP server for production.
Design and operate bounded multi-agent workflows with task decomposition, dependency graphs, ownership, handoff contracts, shared-state controls, approvals, recovery, and synthesis. Use when a task contains genuinely independent workstreams, specialized roles, parallel research or implementation, reviewer-worker loops, or coordination problems that one agent should not execute sequentially.
Design and validate model-facing tool definitions with clear names, action-oriented descriptions, bounded JSON Schema parameters, explicit side effects, safe defaults, idempotency, errors, and realistic tests. Use when creating function-calling tools, MCP tools, agent actions, structured tool inputs, or when a model selects the wrong tool, invents arguments, or causes unsafe side effects.
Plan, execute, document, and retest authorized security assessments of AI agents and multi-agent workflows using safe adversarial cases, synthetic identities, canaries, and evidence-based findings. Use when defining red-team rules of engagement, assessing prompt injection or excessive agency, testing tool and identity boundaries, evaluating memory or cross-agent attacks, scoring a campaign, or verifying remediation in an approved environment.
Threat-model and harden AI agents, RAG systems, assistants, and tool-using workflows against direct, indirect, stored, cross-agent, and multimodal prompt injection. Use when reviewing an agent architecture, isolating untrusted content, constraining tools and egress, protecting secrets, adding injection-focused tests, investigating a suspected injection incident, or documenting residual prompt-injection risk.