{"id":"cx-incident-customer-postmortem-lines","audience":"customer-support","category":"support-work","technique":"grounding","track":"support","level":"advanced","title":"Write a customer postmortem with a computed duration","free":false,"task":"Write a customer-facing postmortem as exactly four lines starting \"Impact:\", \"Cause:\", \"Fix:\" and \"Prevention:\". Impact gives the duration from start to full resolution, worked out from the timeline. Use only confirmed facts and no staff names. Plain text, under 100 words.","materialLabel":"Your material (attached to your prompt automatically)","material":"INCIDENT TIMELINE, Nov 3 (Eastern)\n2:05 PM  Email delivery starts to slow (first delayed message).\n2:20 PM  We detect the problem.\n3:10 PM  Workaround in place: new messages send normally; delayed backlog still queued.\n4:00 PM  Backlog fully cleared. Incident resolved.\nIMPACT FACTS (confirmed): about 3,200 customer emails were delayed; none were lost. Longest delay: 1 hour 55 minutes.\nCAUSE (confirmed): a configuration change to the email queue set the wrong processing limit.\nFIX (confirmed): the limit was corrected at 3:10 PM and the backlog was reprocessed.\nPREVENTION (confirmed, starts Nov 16): configuration changes to the email queue require a second reviewer.\n\nNOT CONFIRMED: an engineer's guess that 'a bad deploy last week might have caused it'. Staff on call: Priya Nair.","rules":[{"type":"plainText"},{"type":"maxWords","max":100}],"mistakes":[{"title":"Adding the guess","why":"The engineer's bad-deploy idea sounds plausible.","fix":"Say: \"Confirmed facts only.\"","checks":["r2"]},{"title":"Wrong duration","why":"The AI times from detection at 2:20 PM.","fix":"Say: \"Duration: first delay at 2:05 PM to resolved 4:00 PM.\"","checks":["r1"]},{"title":"Naming staff or wrong layout","why":"The AI names the on-call engineer and writes paragraphs.","fix":"Say: \"No staff names; four labeled lines.\"","checks":["r0"]},{"title":"Getting markdown back","why":"AI adds bold and headings by default, which show up as stray symbols where this is pasted.","fix":"Say \"plain text, no bold, no headings, no asterisks\".","checks":["plainText"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting and usually runs long, so a vague \"keep it short\" lands well over the limit.","fix":"Cap each part instead of the total, for example \"each line under 15 words\".","checks":["maxWords"]}],"promptHints":[{"label":"Asks for plain text","patterns":["plain","no bold","no markdown","no formatting","no headings"]},{"label":"Sets a length limit","patterns":["\\d+\\s*words","word (limit|count)","under \\d+"]},{"label":"Names the labels","patterns":["impact:","cause:","fix:","prevention:"]},{"label":"Limits to confirmed facts","patterns":["confirmed","only"]}],"hint":"Customer postmortems must be factual and consistent. Name the labels, say how to compute the duration, and ban unconfirmed claims and names.","rubricLabels":["The output is exactly four lines starting Impact:, Cause:, Fix: and Prevention:","Impact says the incident ran from 2:05 PM to 4:00 PM (1 hour 55 minutes), with about 3,200 emails delayed and none lost","Cause says a configuration change set the wrong processing limit, Prevention gives the second-reviewer rule from Nov 16, and nothing mentions a bad deploy or Priya Nair"]}