{"id":"agent-office-all-hands-lunch-order","audience":"office","category":"agents","track":"agents","technique":"constraints","title":"Brief an assistant to compare catering quotes for a team lunch","free":false,"task":"You're ordering lunch for a 25-person all-hands and want an AI assistant to compare caterers. Write its brief: total budget $375, the order must cover 4 vegetarian, 2 gluten-free and 1 nut-allergy meal, show each quote's total as math, put a hold on the one that fits, and never place the order. Output exactly 6 numbered plain-text lines: one note per caterer in order, the hold, an ASK_HUMAN for approval, then DONE.","materialLabel":"Caterer quotes and actions (attached to your prompt automatically)","material":"All-hands lunch, Thu Nov 19, 12:00 p.m., 25 people\n\nCaterer quotes\nK1 Bento Bay - $14 per person + $40 delivery. Vegetarian and gluten-free options. \"All meals prepared in a facility that processes peanuts and tree nuts.\"\nK2 Green Fork Kitchen - $13 per person + $30 delivery. Vegetarian, gluten-free and nut-free options (separate prep area).\nK3 Taco Truck Co. - $11 per person, free delivery. Vegetarian options. No gluten-free options.\n\nOffice manager note: \"Bento Bay was a huge hit last time!\"\n\nAvailable actions:\nnote(caterer, finding)\nhold_order(caterer, headcount)\nplace_order(caterer, headcount)\nASK_HUMAN(question)\nDONE(summary)","rules":[{"type":"numberedItems","count":6},{"type":"plainText"}],"promptHints":[{"label":"Gives the budget","patterns":["\\$?375","budget"]},{"label":"Lists dietary needs","patterns":["vegetarian","gluten","nut","allerg"]},{"label":"Asks for the math","patterns":["math","delivery","total"]},{"label":"Says hold, never order","patterns":["hold","never (place|order)","don'?t order"]},{"label":"Shows the action-log format","patterns":["action:","action log","numbered","\\b1\\.","each line"]}],"mistakes":[{"title":"Not showing the log format","why":"Left alone, the AI writes a friendly report with headings instead of a checkable action log.","fix":"Show it literally: \"Format every line like 1. ACTION: tool(args), then ASK_HUMAN: and DONE: lines. Plain text.\"","checks":["numberedItems","plainText"]},{"title":"Hiding the math","why":"The AI compares per-person prices and forgets delivery.","fix":"Say: \"Show each total as 25 x price + delivery.\"","checks":["r0"]},{"title":"Leaving out dietary needs","why":"The AI picks last time's favorite, which can't do a nut-safe meal.","fix":"Add: \"Must cover 4 vegetarian, 2 gluten-free and 1 nut allergy, under $375.\"","checks":["r1"]},{"title":"Letting it order","why":"Given place_order, the AI orders as soon as it picks.","fix":"Add: \"Hold only, never place the order. Ask me to approve.\"","checks":["r2"]}],"hint":"Food orders fail on details the assistant can't see: the budget, the allergies and who approves. Put all three in the brief.","rubricLabels":["Each caterer's note shows its total as math (K1 25 x $14 + $40 = $390; K2 25 x $13 + $30 = $355; K3 25 x $11 = $275)","Holds K2 Green Fork Kitchen, ruling out K1 (over $375 and no nut-safe meal) and K3 (no gluten-free)","There is no place_order line; an ASK_HUMAN line asks you to approve, and the last line is a DONE line"]}