‹ Engineering recipesEngineering & SRE · Agent Blueprint
Post-mortem prep
Reconstruct yesterday's incident: PagerDuty, #incidents, transcripts, git history, one timeline.
Side by side: token usage, with and without Paperbase
~120k tokens saved · 80% less work for the agentWithout Paperbase
~150k tokens
- Export the PagerDuty timeline for the incident.
- Read the #incidents Slack thread from start to finish.
- Pull the meeting transcript of any war room.
- Search git history for the commits around the window.
- Skim CI/CD logs and dashboards for the same window.
- Manually align timestamps across all sources to build one narrative.
- Reconstruct who said what, when, and what was decided.
With Paperbase
~30k tokens
- Pull the incident timeline as a single Event stream, cross-tool.
- Pull the Interactions around the window: Slack, war room, calls.
- Ask Paperbase to align Events and Interactions to one clock.
- Pull the Decisions made during the incident, with owners and rationale.
- Draft the post-mortem skeleton from the aligned record.
Agent prompt
You are preparing a post-mortem for the incident at {{time_or_incident_id}}.
Using Paperbase:
1. Pull the Event stream for the incident window - alerts, deploys, dashboards - as one timeline.
2. Pull the Interactions in the same window: Slack threads, war room transcript, calls.
3. Align Events and Interactions to a single clock.
4. Pull any Decisions made during the incident, with who made them and why.
5. Note the first signal and the first responder.
Output a post-mortem skeleton:
- Timeline: signal, response, mitigation, resolution, with timestamps.
- Key decisions during the incident, each attributed.
- The gaps: where we were slow, where context was missing.
- Three candidate root causes, grounded in the record.Placeholders in {double_braces} are inputs the agent will ask for at runtime. Give the prompt to any agent connected to Paperbase, and the rest grounds in your own memory graph.