DataAI Nexus

Agents / Skills

Postmortem Write-up

Skill

Writes the blameless post-incident document that is actually worth reading: the honest timeline, the causes past the convenient one, what went well, and actions that would have changed this incident — not a compliance artefact.

Best for

  • Incidents that deserve learning, getting a template instead
  • Making blameless real in writing, not just in the meeting
  • Action lists that would actually have prevented the incident

What you give it

  • The incident's raw material: timeline notes, logs, chat transcripts, the people's accounts

What you get back

  • The document: impact, timeline, causes (plural, honest), what went well, what was luck, actions
  • Blameless done properly: systems and conditions in the language, names only as roles
  • Actions tested against the incident: would each have prevented, shortened or softened THIS one?

How it works

  1. Builds the timeline from evidence with the gaps left visible — the unseen-alert gap is a finding about alerting, not about a person.
  2. Writes causes in system language: conditions, tools, defaults and processes; people appear as roles doing reasonable things with the information they had.
  3. Honours what went well and what was luck separately — luck is a gap wearing a medal.
  4. Filters actions by the counterfactual test: each must plausibly have prevented, shortened or softened this incident, with an owner and a date.

Example

You: Write up Tuesday's 40-minute checkout outage from this chat log and these notes.

Result: The document: impact quantified (orders lost, estimated revenue, support volume), the timeline with the uncomfortable 14-minute gap where the alert fired unseen (named as a finding, not a person's shame), three causes (the config change, the single-channel alert, the rollback that needed a credential only two people hold), two things that went well, one acknowledged luck (the senior was awake), and four actions — each tested: 'would this have changed Tuesday?' Three yes, one cut for being ritual.

Limits — please read

  • The document is as honest as the inputs; a sanitised chat log yields a sanitised truth — it will note where accounts conflict.
  • Blameless language cannot fix a blameful culture, but it can refuse to feed it.
  • Action follow-through lives in your tracker; the document links and the review checks.