RCA, Runbook, and Postmortem Prompt Templates
Use seven ready prompt templates for RCA, runbooks, stakeholder updates, and postmortems, filled from real incident data and chained through an incident.
What You'll Learn
Understanding Why Templates Beat Ad Hoc Prompts
Where acme-shop stands By now acme-shop can detect the connection-pool incident, find it with the manual workflow from RCA and Production Debugging...
Setting Up the Template Lab
What the lab needs Everything lives in aiops-lab/prompts/, next to the kit's llm.py and its runbooks/ folder. Every model call goes through llm.py.
Using the RCA Templates
Template 1: RCA from logs and metrics Use this when an alert fires and you have a few log lines and some metrics.
Using the Runbook Templates
Template 4: Write a runbook for a new alert Use this when you add an alert and need a runbook for it. Name the environment exactly.
Using the Incident Update and Postmortem Templates
Template 6: Stakeholder update Use this during a live incident to brief people who are not technical.
Chaining Templates Through an Incident
How the templates hand off In a real incident you do not use one template. The output of one becomes the input of the next.
Skills You'll Master
Curriculum Index9 topics
Understanding Why Templates Beat Ad Hoc Prompts
Where acme-shop stands By now acme-shop can detect the connection-pool incident, find it with the manual workflow from...
Setting Up the Template Lab
What the lab needs Everything lives in aiops-lab/prompts/, next to the kit's llm.py and its runbooks/ folder.
Using the RCA Templates
Template 1: RCA from logs and metrics Use this when an alert fires and you have a few log lines and some metrics.
Using the Runbook Templates
Template 4: Write a runbook for a new alert Use this when you add an alert and need a runbook for it.
Using the Incident Update and Postmortem Templates
Template 6: Stakeholder update Use this during a live incident to brief people who are not technical.
Chaining Templates Through an Incident
How the templates hand off In a real incident you do not use one template.
Running the Hands-On Lab
Cost and setup 📌 Remember: With the default local Ollama model through llm.py, this lab costs nothing, and the...
What You Built and What Comes Next
What you built You built a reusable prompt library for the incident workflow: seven templates, a loader that fills them...
Quick Reference and Common Mistakes
Quick reference Common mistakes Filling a template by hand under pressure leaves fields out.
Career Impact
Roles that use the skills in this module.
- High Demand
AIOps Engineer
₹18L - ₹35L a year
- High Demand
SRE
₹20L - ₹38L a year
- Growing
Platform Engineer
₹15L - ₹28L a year
Next Modules
Related Guides
Practice on the Coding Sheet
Not a software engineer sheet. Every problem comes from real DevOps, SRE, Platform and Cloud interviews, from your first script to a system you build yourself.
Open the Coding SheetFrequently Asked Questions
During an incident nobody has time to remember which details a good prompt needs. A template makes the right inputs the same every time, so the model gets complete data and you can compare results from one incident to the next.
Yes, if you ask for JSON and validate it against a schema before anything acts on it. This module shows a JSON version of the RCA template with a retry and a safe fallback to a human.
Fill it from a real or replayed incident and run automatic checks on the reply: does the root cause name the right component, does the update contain jargon, does the postmortem have every section. A template that fails its checks needs a rewrite.
Not from these templates. They ask for read-only checks that a person runs. Allow lists and approval gates for acting on model output are covered in AI Guardrails and Output Validation.