{"id":"fpa-budget-guardrail-checks","audience":"fpa","category":"finance-work","track":"finance","technique":"constraints","level":"beginner","title":"Check a department budget submission against four guardrails","free":false,"task":"You're screening FY27 budget submissions. Write a prompt for plain text with four lines starting \"Growth cap:\", \"New heads:\", \"One-time split:\" and \"Assumptions:\", each saying Pass or Fail and the evidence with the math, then a last line \"Overall:\" that says Ready only if all four pass, otherwise Not ready. Under 90 words. No bullets or numbering.","materialLabel":"Guardrails and one department's submission (attached to your prompt automatically)","material":"FY27 BUDGET GUARDRAILS\n1. Growth cap: total department budget may grow at most 8% over FY26.\n2. New heads: every new position requested needs an approved requisition number.\n3. One-time split: the submission shows one-time and recurring amounts separately.\n4. Assumptions: the submission includes a documented assumptions tab.\n\nSUBMISSION - Customer Success department\nFY26 budget: $1,250,000. FY27 request: $1,375,000.\nNew positions requested: 3 (CS-11 and CS-12 have approved requisition numbers; the third has none yet).\nThe request includes $60,000 one-time (training platform setup) and $1,315,000 recurring, shown on separate lines.\nNo assumptions tab was attached. The department head wrote: 'Assumptions are in my head and obvious.'","rules":[{"type":"plainText"},{"type":"maxWords","max":90},{"type":"maxBullets","max":0}],"promptHints":[{"label":"Names the five lines","patterns":["growth cap","new heads","one-time split","assumptions","overall"]},{"label":"Asks for Pass or Fail","patterns":["pass or fail"]},{"label":"Shows the math","patterns":["math","growth"]},{"label":"States the overall rule","patterns":["ready only if","not ready"]},{"label":"Caps length","patterns":["\\d+\\s*words","plain"]}],"mistakes":[{"title":"Accepting the assumptions claim","why":"The AI passes assumptions because the head says they're obvious.","fix":"Add: \"Judge by what was submitted, not by what the head says.\"","checks":["r2"]},{"title":"No growth math","why":"The AI says 'a bit above the cap'.","fix":"Say: \"Show the growth math: increase / FY26 budget.\"","checks":["r0"]},{"title":"Wrong overall","why":"The AI says Ready because most checks pass.","fix":"Add: '\"Ready\" only if all four pass.'","checks":["r2"]},{"title":"Markdown extras","why":"The AI adds bold labels, headings and a closing note.","fix":"Say: \"Plain text, no bold. Output only the lines asked for.\"","checks":["plainText","maxBullets","r1"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting and runs past word limits.","fix":"Ask for fewer words than the real cap and keep each line short.","checks":["maxWords"]}],"hint":"A guardrail check is pass or fail with evidence. Name the four tests and the overall rule.","rubricLabels":["Growth cap line says Fail: $1,375,000 vs $1,250,000 is +$125,000, which is 10.0% and above the 8% cap ($1,350,000 maximum)","New heads line says Fail (3 requested, 2 approved requisitions) and One-time split line says Pass ($60,000 one-time, $1,315,000 recurring)","Assumptions line says Fail (no assumptions tab) and the last line is \"Overall: Not ready\""]}