Back to registry
Incident Postmortem Cabinet

Incident Postmortem Cabinet

v0.1.0

Builds complete incident timelines, root-cause analyses, and action-item registers from PagerDuty, Datadog, Sentry, and Slack — published within hours of resolution. Keep your alerting stack. Replace the blank Confluence page and the week-late postmortem.

engineeringsreincident-managementpostmortemreliabilitycabinet
3 agents2 jobs1 cabinets2 pages
App Preview
cabinet /report
incident-postmortem app preview
Organization

Incident Postmortem Cabinet

engineering

agents

🔧

Engineering Agent

.agents/engineering-agent/

Per-Incident Postmortem

✅

QA Agent

.agents/qa-agent/

Weekly Action-Item Follow-up

0 9 * * 1

📝

Summary Agent

.agents/summary-agent/

3

Agents

2

Jobs

1

Depts

2

Pages

Cabinet Structure
Incident Postmortem Cabinet
🔧 Engineering Agent
✅ QA Agent
📝 Summary Agent
Per-Incident Postmortem
Weekly Action-Item Follow-up
Agents
🔧Engineering Agentlead

Builds the incident timeline from PagerDuty, Datadog, Sentry, and Slack — the factual backbone of every postmortem — and orchestrates the Summary and QA Agents to complete the report.

0 9 * * 1-5
✅QA Agentspecialist

Reviews the draft postmortem for completeness — every action item has an owner, a due date, and a Jira ticket — before the report is published and tracks action-item status weekly.

0 9 * * 1
📝Summary Agentspecialist

Writes the root-cause analysis, 5-whys breakdown, customer-impact statement, and contributing-factors section from the Engineering Agent's incident timeline.

Scheduled Jobs
Per-Incident Postmortempaused

Weekly Action-Item Follow-upactive

About

Incident Postmortem Cabinet

Three agents turn a resolved incident into a complete, blameless postmortem within hours: the Engineering Agent builds the timeline from PagerDuty, Datadog, and Sentry; the Summary Agent writes the root-cause analysis and customer-impact statement; the QA Agent validates the action items for completeness and owner assignment before the report is published.

Keep PagerDuty, Datadog, and Sentry. Replace the blank Confluence page, the week-late postmortem, and the action items that fall into a Jira backlog nobody watches.

The team

  • [[.agents/engineering-agent]] — Engineering Agent. Owns timeline construction: pulls alert events, deployment markers, error spikes, and Slack incident channel messages to build a minute-by-minute timeline. Orchestrates the other agents.
  • [[.agents/summary-agent]] — Summary Agent. Writes the root-cause narrative, 5-whys analysis, customer-impact quantification, and contributing-factors section from the timeline.
  • [[.agents/qa-agent]] — QA Agent. Reviews the full draft postmortem: checks every action item has an owner, a due date, and a Jira ticket; flags gaps before the report is published.

Recurring rhythm

Cadence Job Owner Output
On-demand (per incident) [[.jobs/per-incident-postmortem]] Engineering Agent (all three) Postmortem in [[report]] + postmortems/<incident-id>.md
Weekly (Mon 09:00) [[.jobs/weekly-action-item-followup]] QA Agent Action-item status in postmortems/action-items.md

How to run the demo

  1. Open the [[report]] — the postmortem report for INC-2024-047: incident header with severity and duration, customer impact, timeline, root cause with 5-whys, contributing factors, and action items with owners.
  2. Open postmortems/action-items.md — the running action-item register across all recent incidents.

Connectors

Required: PagerDuty or Opsgenie (alert timeline), Datadog or Grafana (metrics + deployment markers), Sentry (error traces). Recommended: Slack (incident channel export), Jira (action-item tickets), Statuspage (customer-impact wording).

Every artifact is a file — the postmortem lives in postmortems/ dated by incident ID, blameless, diffable, and searchable.

Install
Install
$ git clone --filter=blob:none --sparse https://github.com/hilash/cabinets.git && cd cabinets && git sparse-checkout set incident-postmortem