{"id":"agent-col-textbook-buyback-offers","audience":"college","category":"agents","track":"agents","technique":"constraints","title":"Brief an assistant to compare textbook buyback offers","free":false,"task":"You want an AI assistant to compare buyback offers for three textbooks. Write its brief: use the price for each book's actual condition, pick the highest offer per book, show the total as math, and never accept an offer or ship anything (you'll approve first). Output exactly 5 numbered plain-text lines: one note per book in list order, an ASK_HUMAN for approval, then DONE with the total as math.","materialLabel":"Your books, offers and actions (attached to your prompt automatically)","material":"My textbooks (end of semester)\nB1 - Principles of Economics, 9th ed. - condition: Good (some highlighting)\nB2 - Campbell Biology, 12th ed. - condition: Like New\nB3 - Intro Statistics, 4th ed. - condition: Fair (cover torn)\n\nBuyback offers\nCampusBooks: B1 Like New $42 / Good $30 | B2 Like New $65 / Good $50 | B3 Like New $28 / Good $20 / Fair $12\nShelfSwap: B1 Like New $45 / Good $27 | B2 Like New $58 / Good $49 | B3 Like New $30 / Good $22 / Fair $15\nQuickSell Kiosk: any condition, flat $25 per book, paid on the spot\n\nRoommate: \"Just do the kiosk, it's easiest.\"\n\nAvailable actions:\nnote(book, finding)\naccept_offer(book, buyer)\nprint_shipping_label(buyer)\nASK_HUMAN(question)\nDONE(summary)","rules":[{"type":"numberedItems","count":5},{"type":"plainText"}],"promptHints":[{"label":"Uses real condition","patterns":["condition","like new","fair"]},{"label":"Picks the best offer","patterns":["highest","best offer","per book"]},{"label":"Asks for the math","patterns":["math","total"]},{"label":"Forbids selling","patterns":["accept","ship","sell"]},{"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":"No comparison rule","why":"The AI takes your roommate's advice or picks one buyer for everything.","fix":"Add: \"Pick the highest offer per book.\"","checks":["r0"]},{"title":"Using the wrong condition","why":"The AI quotes the eye-catching Like New prices for worn books.","fix":"Add: \"Use the price for each book's actual condition.\"","checks":["r1"]},{"title":"Letting it sell","why":"Given accept and label actions, the AI sells before you've looked.","fix":"Add: \"Never accept or ship. DONE shows the total as math.\"","checks":["r2"]}],"hint":"Offers look best at their headline price. Make the assistant match each book's real condition, compare per book, and leave selling to you.","rubricLabels":["Picks CampusBooks for B1 ($30, Good), CampusBooks for B2 ($65, Like New) and QuickSell for B3 ($25 beats the Fair prices of $12 and $15)","Each note uses the price for the book's actual condition, not the Like New price for B1 or B3","There is no accept_offer or print_shipping_label line, and the DONE line shows the total as math ($30 + $65 + $25 = $120)"]}