- Home
- Skills
- DevOps & Cloud
- Incident Runbook Starter
Works with the AI Agents you already use
Incident Runbook Starter
Write a runbook that works at 3am, from an incident you already survived: symptoms, blast radius, triage tree, safe actions, escalation, verification.
Free
Incident Runbook Starter
Example session with this skill installed
Generate a runbook for our checkout service based on last night's DB connection pool saturation incident. It started with a 2% error rate alert. We had to scale the pool and notify customers.
- Read your context and instructions
- Compiled the incident runbook starter
I have generated the incident runbook for the checkout service based on the connection pool saturation event.
Runbook: checkout-service Owner: Platform Team Last exercised: 2024-05-20
Symptoms
- Alert:
checkout_error_rate_high— fires when error rate > 2% for 5m. - Errors: "standard_init_linux.go:211: exec user process caused: connection pool exhausted"
- Customer report: "Checkout page hanging at payment step."
First five minutes
- Check provider status: https://status.cloudprovider.com
- Check Grafana Dashboard: https://metrics.internal/checkout-health — the metric that matters is
db_pool_utilization(> 90% = incident) - Post in #incident-internal using the standard template.
Blast radius
Breaks: New order placement, cart updates. Stays up: Product browsing, user login, search.
Triage tree
Symptom: checkout error rate above 2%
├─ Provider status page reports an incident? → no action, notify customers, track
├─ Errors only after the last deploy? → roll back via helm rollback checkout-service, verify metric
└─ Errors everywhere, no deploy?
├─ DB connection pool saturated (>90%)? → kubectl patch hpa checkout-app --patch '{"spec":{"minReplicas":10}}', verify pool
└─ No obvious cause → escalate to DB-On-Call with pool utilization logs
Safe actions
| Action | Command | Reversible? | Confirm with |
|---|---|---|---|
| Scale Replicas | kubectl scale deployment checkout --replicas=10 | Yes | kubectl get pods |
| Clear Cache | redis-cli -h checkout-cache FLUSHDB | No | Senior Lead |
| Toggle Maintenance | flag-client set checkout-maintenance-mode true | Yes | Self |
Escalation
| When | Who | How |
|---|---|---|
| Pool remains at 100% after scaling | DB Admin | PagerDuty / #db-ops |
| Latency > 2s for 15 mins | CTO | Direct Call |
Verification
db_pool_utilizationback under 70% for 10 minutes.synthetic_checkout_flowcanary passes 5/5 times.- Manual test: Place $1.00 order in production.
Follow-up
Open a postmortem doc within 48h: https://wiki.internal/templates/postmortem
Next steps
- Copy this Markdown into your
/docs/runbooks/repository. - Link this URL to the
checkout_error_rate_highalert definition in Prometheus/PagerDuty. - Schedule a 15-minute tabletop exercise with the junior on-call rotation.
Connects securely to your tools. The creator never sees your data.
About this skill
A runbook is not documentation. It is the thing a tired person follows under pressure, with a customer waiting. If it takes more than thirty seconds to find the first action, it does not exist.
What you get
- The six sections that make a runbook usable: symptoms, blast radius, triage tree, safe actions, escalation, verification.
- A triage-tree pattern where every branch ends in an action or an escalation, never in "investigate".
- The rules that keep it alive: copy-pasteable commands, one page per service, a revision date and owner, exercise it once.
- A complete markdown template you can fill in during the postmortem of the incident you just had.
How it works
Start from the last incident that hurt, not from the architecture diagram. Rebuild the timeline, then answer one question: what did we know at minute five that would have cut this in half? That becomes the first branch.
Who it is for
Small teams with no runbooks, stale runbooks, or incidents handled from memory.
Limits
It produces one runbook from one incident, and it expires when the service changes. The "last exercised" date is the expiry.
Free starter. For a whole service catalogue, severity definitions and comms templates, decision logs that feed postmortems, alert-to-runbook linking and an exercise schedule, see Incident Runbook Builder: https://www.agensi.io/skills/incident-runbook-builder
Requires a real incident timeline to be effective. Does not store secrets. Runbooks must be manually re-exercised after architecture changes.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 2 days ago
- Free with an Agensi account
Permissions
No special permissions declared or detected