| name | incident |
| description | IT incident response workflow. This skill should be used when the user asks to "log an incident", "handle an incident", "create incident report", "document an outage", "incident response", "service is down", "major incident", or describes a service disruption requiring structured response. Also use when the user mentions SRE concepts like error budget impact, blameless review, or AIOps-detected incidents. Guides through the full ITIL v5 incident lifecycle: detect, classify, investigate, resolve, communicate, and close. |
| allowed-tools | ["Read","Write","Edit","Glob","Grep","Bash","WebSearch","WebFetch","AskUserQuestion"] |
| argument-hint | <description of incident or symptoms> |
| user-invocable | true |
Incident Response Workflow
Structured incident response following ITIL 4 Incident Management. Produces a complete incident report as a local markdown file.
Related Skills (auto-loaded)
- itil — priority classification (P1-P4), incident management process, escalation matrix
- comms — stakeholder notification templates, severity-based tone guidance
- done-ops — verification gate before closing the incident
Output
File: {OUTPUT_DIR}/incident-{YYYY-MM-DD}-{slug}.md
- slug: lowercase kebab-case summary, e.g.
email-outbound-failure, vpn-india-outage
- Incident ID (
INC-{YYYYMMDD}-{NNN}): {NNN} is a daily sequence number starting at 001. If the organization uses a ticketing system (ServiceNow, Jira SM), use that system's ID instead.
Default output directory: ./incidents/. If a docs/incidents/ directory exists in the project, use that instead. Ask the user if neither exists. A user-specified path takes precedence over all defaults.
All incident documents are written in English. Before writing to file, preview the content for the user in their preferred language.
Workflow
Step 1: Gather Information
Collect the following from the user (use AskUserQuestion for missing items):
| Field | Required | Example |
|---|
| What is happening? | Yes | "Email service is not sending outbound messages" |
| When did it start? | Yes | "Noticed at 09:15 today" / "Alerts fired at 14:30" |
| Who/what is affected? | Yes | "All users in HQ" / "Customer-facing API" |
| Severity indicators | Yes | Is there a workaround? How many users? Revenue impact? |
| What has been tried? | No | "Restarted the mail service, no change" |
Step 2: Classify (SEV then P)
First — assign Severity (objective impact scope):
| Level | Scope | Definition |
|---|
| SEV-1 | All users / all regions | Core service completely down, no workaround |
| SEV-2 | Many users or major function | Significant degradation, partial workaround |
| SEV-3 | Subset of users, non-critical | Limited impact, workaround available |
| SEV-4 | Single user or negligible | Cosmetic, proactive, informational |
Then — assign Priority (handling order, Impact × Urgency):
| High Urgency | Medium Urgency | Low Urgency |
|---|
| High Impact | P1 Critical | P2 High | P3 Medium |
| Medium Impact | P2 High | P3 Medium | P4 Low |
| Low Impact | P3 Medium | P4 Low | P4 Low |
Present both SEV and P to the user for confirmation. MUST use the exact format: SEV-1, SEV-2, P1, P2 (not "Critical" or "High" alone).
If SEV-1 or P1/P2: Major Incident procedures apply (see itil skill reference references/incident-management.md).
Step 3: Document Timeline
Create the incident report file with the initial entry. MUST follow this exact structure — do not rearrange or rename sections:
# Incident Report: {TITLE}
| Field | Value |
|---|---|
| Incident ID | INC-{YYYYMMDD}-{NNN} |
| Severity | SEV-{1/2/3/4} |
| Priority | P{1/2/3/4} |
| Status | Investigating |
| Reported by | {NAME/SYSTEM} |
| Start time | {YYYY-MM-DD HH:MM TZ} |
| Affected services | {LIST} |
| Affected users | {SCOPE} |
## Impact Summary
{One paragraph describing what users are experiencing.}
## Timeline
| Time | Event |
|---|---|
| {HH:MM} | {Event description} |
## Investigation
{Findings so far, diagnostics performed, hypotheses.}
## Resolution
{To be completed upon resolution.}
## Follow-up Actions
- [ ] {Action items identified during the incident}
Step 4: Draft Communications
Using the comms skill, draft the appropriate notifications:
- Initial notification — based on priority level and audience
- Present the draft to the user for review before including in the report
For P1/P2: remind the user about the update cadence (P1: every 15-30 min, P2: every 30-60 min).
If a status page update is also needed, coordinate with the /statuspage skill. The incident report contains full technical detail; the status page update contains only impact-focused, user-safe information.
Step 5: Track Progress
As the user provides updates:
- Append to the Timeline section
- Update the Status field (Investigating → Identified → Monitoring → Resolved)
- Draft follow-up communications as needed
- Add to Investigation section with findings
Step 6: Resolution
When the user indicates the issue is resolved:
- Load done-ops verification gate
- Walk through the incident closure checklist:
- Service restored and verified?
- User/reporter confirmation obtained?
- Monitoring stable (no recurring alerts)?
- Documentation complete?
- Problem record needed? (recurring incident or unknown root cause)
- Update the incident report:
- Set Status to Resolved
- Record resolution time
- Fill in the Resolution section
- Calculate total duration
- Draft the resolution notification using comms templates
Step 7: Close and Follow-up
Complete the incident report:
- Record all follow-up actions with owners and due dates
- Note if a Post-Incident Review is needed (mandatory for P1/P2, recommended for P3)
- Recommend creating a Problem record if root cause is unknown
- Present the final report to the user for review
SRE Postmortem (for SEV-1 and SEV-2): If this was a SEV-1 or SEV-2 incident, a traditional PIR may not be sufficient. Prompt the user:
"This was a {SEV-1/SEV-2} incident. Consider running a blameless postmortem using /postmortem — it produces a more systematic analysis with contributing factors, what went well, and SMART action items. Use /problem if root cause is unknown and the incident is likely to recur."
Error budget: If the organization tracks SLOs, note the estimated error budget impact in the follow-up section.
Gotchas
- Always use the user-facing service name in communications, not internal system names
- Never include IP addresses, hostnames, or credentials in the incident report
- For P1 incidents, do not wait for complete information before drafting the initial notification — speed matters
- Resolution time = time from detection to confirmed restoration, not when the ticket is closed
- If the incident was caused by a recent change, link the change request ID in the report
- Ask the user whether this incident needs to follow their organization's ticketing system (ServiceNow, Jira Service Management, etc.) in addition to the local markdown report