{"id":"cs-escalation-draft-five-checks-review","audience":"managers","category":"support-work","technique":"verification","track":"support","level":"advanced","title":"Review a CSM's escalation draft against five checks","free":false,"task":"Review the draft escalation email against the five checks. Output five lines \"C1:\" to \"C5:\", each \"Pass\" or \"Fail\" followed by a dash and an exact quote from the draft that proves it. Then a last line: \"Send as is:\" with Yes or No. Plain text, under 100 words.","materialLabel":"Your material (attached to your prompt automatically)","material":"CHECKS\nC1: Blames no named person or team.\nC2: Promises no fix date.\nC3: Includes no revenue or ARR figure.\nC4: Contains exactly one ask.\nC5: Says what has already been tried.\n\nDRAFT (CSM Sam to Maria, VP Customer Success)\nSubject: URGENT - Halden Corp is furious\n\nHi Maria, Halden (ARR $410k) is very unhappy. Their reports have failed since Oct 27. We already re-ran the sync on Oct 30, but the failures came back. Support (Jens) ignored their P1 for three days and Product keeps missing dates. I promise we will have it fixed by Friday. Can you speak to their CEO, review the contract and help me staff a dedicated engineer? Also let me know about the credit. Thanks, Sam","rules":[{"type":"plainText"},{"type":"maxWords","max":100}],"mistakes":[{"title":"Invented quotes","why":"The AI paraphrases instead of quoting the draft.","fix":"Say: \"Quote the exact words from the draft.\"","checks":["r0","r1"]},{"title":"Passing a multi-ask email","why":"The AI treats the questions as one request for help.","fix":"Say: \"C4 fails if there is more than one ask.\"","checks":["r2"]},{"title":"Skipping the verdict","why":"The AI gives feedback in prose.","fix":"Say: \"End with Send as is: Yes or No.\"","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":"The AI runs long and cannot count its own words.","fix":"Cap each part (\"each line under 12 words\") and add \"under N words in total\".","checks":["maxWords"]}],"promptHints":[{"label":"Gives the check labels","patterns":["c1","c5"]},{"label":"Asks for exact quotes","patterns":["exact quote","quote","verbatim"]},{"label":"Pass or Fail with a dash","patterns":["pass","fail","dash"]},{"label":"Adds the verdict line","patterns":["send as is"]},{"label":"Asks for plain text","patterns":["plain","no bold","no markdown","no formatting","no headings"]},{"label":"Sets a word cap","patterns":["\\d+\\s*words","under \\d+","word"]}],"hint":"To grade a draft reliably, give fixed checks and require exact quotes as evidence.","mission":"Check a CSM's escalation draft against all five checks.","musts":["Pass or Fail each","Exact quote as proof","Final Send as is verdict"],"rubricLabels":["The output has five lines C1: to C5: each starting with Pass or Fail, a dash and a quote, followed by a last line starting \"Send as is:\"","C1 is Fail quoting the blame of Support or Jens, C2 is Fail quoting \"fixed by Friday\", and C3 is Fail quoting \"$410k\"","C4 is Fail (more than one ask, such as speaking to the CEO, reviewing the contract, staffing an engineer or the credit), C5 is Pass quoting the re-run of the sync on Oct 30, and the last line is \"Send as is: No\""]}