| name | triage-tickets |
| description | Triage support tickets or issues — categorise, assign severity, identify routing, and flag patterns. |
| argument-hint | [ticket content, issue list, or path to issues] |
| user-invocable | true |
| allowed-tools | Read, Bash, Glob, Grep |
Triage $ARGUMENTS using the mandatory process below. Every step is required — do not skip any.
Step 1 — Ingest and normalise
Read every ticket. For each, extract:
- Ticket ID (or assign a sequential one if absent)
- Reporter (name, account, plan tier if available)
- Raw summary — one sentence capturing what the user actually said
- Reproduction signal — does the ticket contain steps to reproduce, error messages, or screenshots?
If tickets are in a file or directory, use Read, Glob, or Grep to pull them. If they are inline, parse directly.
Step 2 — Classify each ticket
Apply ALL of the following dimensions to every ticket:
Category
| Category | Use when |
|---|
| Bug | Something that worked before is now broken, or behaves contrary to documentation |
| Feature request | User wants new functionality that does not exist |
| How-to question | User is trying to accomplish something and needs guidance |
| Account / billing | Login, permissions, invoices, plan changes |
| Integration issue | Problem occurs at the boundary with a third-party system |
| Performance | Slowness, timeouts, resource exhaustion — system works but poorly |
| Data issue | Missing, corrupted, or incorrect data |
| Complaint | Expressing dissatisfaction without a specific technical issue |
Severity
| Severity | Definition | Response target | Examples |
|---|
| Critical | System down, data loss, security breach, or complete blocker for multiple users | Acknowledge within 1 hour, update every 2 hours | Production outage, data deletion, auth bypass |
| High | Major feature broken for a user, no workaround available | Acknowledge within 4 hours, update daily | Cannot export data, payment processing fails |
| Medium | Feature degraded but workaround exists, or issue affects a non-critical path | Acknowledge within 1 business day | Slow report generation, UI glitch on one browser |
| Low | Cosmetic issue, minor inconvenience, or question with self-serve answer | Acknowledge within 2 business days | Typo in UI, color preference, "how do I?" |
Severity escalation rules:
- Enterprise or high-tier customer reporting → raise severity by one level
- Issue mentions data loss or security → automatically Critical
- Issue blocks a user from core workflow → at least High
Routing
| Route to | When |
|---|
| Engineering — bugs | Confirmed bug with reproduction steps |
| Engineering — infrastructure | Performance, uptime, scaling issues |
| Product | Feature requests, UX complaints with 3+ occurrences |
| Documentation | How-to questions answerable by better docs |
| Support (self) | Account, billing, known-issue with documented workaround |
| Security | Any mention of unauthorised access, data exposure, vulnerability |
Workaround
For every ticket, answer: Is there a known workaround?
- If yes: state the workaround in 1-2 sentences, cite the source (KB article, prior ticket)
- If no: state "No known workaround" explicitly
- If uncertain: state "Workaround unknown — needs investigation"
Step 3 — Detect patterns
After classifying all tickets, scan for patterns:
- Duplicate cluster: 2+ tickets describing the same root issue → group them, note the count
- Escalation trigger: 3+ tickets on the same issue → flag for immediate escalation to product/engineering with this template:
PATTERN ESCALATION
Issue: [one-line description]
Ticket count: [N]
Affected users: [list or count]
Severity: [highest severity among the cluster]
Sample ticket IDs: [up to 5]
First reported: [date of earliest ticket]
Workaround available: [yes/no]
Recommended action: [suggested next step]
- Regression signal: issue appeared after a known deploy or release → flag with the release version
- Trend direction: is this issue increasing, stable, or decreasing over the batch?
Step 4 — Generate bug reports for engineering
For every ticket routed to engineering, produce a structured bug report:
## Bug Report: [title]
**Source tickets**: [IDs]
**Severity**: [level]
**Category**: [bug / performance / infrastructure]
### Description
[What is happening, in plain language]
### Steps to reproduce
1. [Step]
2. [Step]
3. [Step]
### Expected behaviour
[What should happen]
### Actual behaviour
[What happens instead]
### Environment
- Product version: [if known]
- Browser/OS/client: [if known]
- Account/plan: [if known]
### Supporting evidence
[Error messages, screenshots, logs — quote directly from the ticket]
### Workaround
[If any]
If the ticket lacks reproduction steps, state what is missing and recommend the support agent ask the user for specific information.
Step 5 — Output
Triage table
Present ALL tickets in a single table, sorted by severity (Critical first), then by category:
| Ticket ID | Summary | Category | Severity | Route to | Workaround | Pattern cluster |
|---|
| ... | ... | ... | ... | ... | ... | ... |
Pattern summary
After the table, list:
- Escalations triggered — any pattern clusters hitting the 3+ threshold
- Regression candidates — issues potentially linked to recent changes
- Workaround gaps — high/critical tickets with no known workaround (these need KB articles)
Metrics
- Total tickets triaged: [N]
- By severity: Critical [N], High [N], Medium [N], Low [N]
- By routing: Engineering [N], Product [N], Docs [N], Support [N], Security [N]
- Patterns detected: [N] clusters covering [N] tickets
Rules
- Never downplay a user's reported severity without evidence. If a user says "this is urgent," treat it as at least High until investigation says otherwise.
- When in doubt about category, pick the one that gets the ticket to the right team faster.
- Always preserve the user's original language in the summary — do not sanitise or corporate-speak their words.
- If a ticket contains multiple issues, split it into separate entries in the triage table.
- Every Critical and High ticket MUST have a recommended next action, not just a routing destination.
Related Skills
/support:feedback-synthesis — after triaging a batch, synthesise recurring themes into product feedback.
/support:write-kb-article — when triage reveals a common user issue, write a KB article to reduce future ticket volume.