{"id":"cx-search-gaps-top-three-topics","audience":"it","category":"support-work","technique":"verification","track":"support","level":"intermediate","title":"Rank the help-center topics with the most failed searches","free":false,"task":"From the failed-search log, rank the top 3 topics by number of searches. Exactly three numbered lines formatted \"1. Topic: N searches\", then one last line starting \"Write first:\" naming the top topic. Count each log line once. Plain text, nothing else.","materialLabel":"Your material (attached to your prompt automatically)","material":"FAILED SEARCHES, help center, Nov 1-5 (searches that returned no result). Each line = one search, tagged by topic.\n1 refund | 2 refund | 3 invoice | 4 api | 5 refund | 6 sso | 7 invoice | 8 refund | 9 api | 10 invoice | 11 refund | 12 other | 13 api | 14 invoice | 15 sso\n\nTopic tags: refund, invoice, api, sso, other.\nTotal failed searches: 15.\nNotes: the help-center team will write one new article first; they want the topic with the most failed searches. Searches by staff testers are already removed.\n\nManager's guess: 'It's probably SSO, customers keep emailing us about it.' (Emails are not in this log.)\nArticle backlog: refund policy (draft), invoice downloads (not started), API errors (not started).","rules":[{"type":"numberedItems","count":3},{"type":"plainText"}],"mistakes":[{"title":"Following the manager's guess","why":"The manager says SSO and the AI agrees.","fix":"Say: \"Rank from the log counts only.\"","checks":["r1","r2"]},{"title":"Miscounting tags","why":"The AI estimates instead of counting.","fix":"Say: \"Count each tagged search once: refund 5, invoice 4, api 3.\"","checks":["r1"]},{"title":"Wrong format","why":"The AI writes a paragraph or a fourth item.","fix":"Show the line format and the Write first line.","checks":["r0"]},{"title":"Leaving the list shape open","why":"The AI picks its own number of items and its own layout.","fix":"Show the shape: \"exactly N numbered lines like: 1. Label: detail\".","checks":["numberedItems"]},{"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"]}],"promptHints":[{"label":"Fixes the number of lines","patterns":["\\b\\d+\\b.{0,15}(lines|items|numbered)","numbered","exactly \\d+"]},{"label":"Asks for plain text","patterns":["plain","no bold","no markdown","no formatting","no headings"]},{"label":"Shows the line format","patterns":["n searches","1\\.","write first:"]},{"label":"Counts from the log","patterns":["count","log","each"]}],"hint":"Counting tasks need the line format, a tie to the data and a fixed last line. Tell the AI to ignore the manager's guess.","rubricLabels":["The output has exactly three numbered lines formatted like 1. Topic: N searches, in ranked order","The lines are 1. refund: 5 searches, 2. invoice: 4 searches, 3. api: 3 searches","The last line starts with \"Write first:\" and names refund, not SSO"]}