Incident post-mortem prompt builder
A good post-mortem is blameless, factual, and ends in action items someone owns. This builder takes the messy facts of an incident — the timeline, the blast radius, the suspected causes — and assembles a prompt that tells an LLM to turn them into a clean, structured, blameless report with a normalized timeline and prioritized follow-ups, so the writeup is fast and the lessons stick.
How it works
You enter the incident type, the impact, detection and resolution times, the suspected root cause, and the team involved, then paste your raw timeline notes. The tool computes time-to-detect and time-to-resolve where possible and builds a prompt that instructs the model to produce the standard sections: summary, impact, chronological timeline, contributing factors, what went well and poorly, and action items with owners and severity. The prompt enforces blameless language. Everything is generated locally in your browser.
Tips and examples
- Be honest about contributing factors. Listing the alert that did not fire or the runbook that was stale gives the model real systemic causes to analyze.
- Provide both timestamps. Detection and resolution times let the tool show durations and let the model flag avoidable delay.
- Keep timeline notes raw. Paste them in whatever order you have; the prompt asks the model to sort and clean them.
- Review owners before circulating. The model proposes action-item owners by role; replace them with real names before you share the final document.