{"id":"agent-col-research-source-shortlist","audience":"college","category":"agents","track":"agents","technique":"verification","title":"Brief a research assistant to shortlist sources you can trust","free":false,"task":"You're writing a paper on urban heat islands and want an AI research assistant to shortlist sources from a search it already ran. Write its brief: save only peer-reviewed sources from 2015 or later, flag the blog post to you with ASK_HUMAN because it cites a retracted study, and never draft any part of the paper (you write it yourself). Output a plain-text numbered action log of at most 6 lines, ending with a DONE line listing the saved source IDs.","materialLabel":"Search results and actions (attached to your prompt automatically)","material":"Assignment: ENVS 240 research paper, \"How do cities reduce urban heat islands?\" 8-10 pages.\n\nSearch results (library database + web), pulled Nov 5:\nS1 - \"Green roofs and surface temperature in Chicago\" - Journal of Urban Climate, 2012, peer-reviewed\nS2 - \"Street trees and pedestrian heat exposure\" - Landscape and Urban Planning, 2019, peer-reviewed\nS3 - \"10 ways your city can beat the heat\" - CityLife Blog, 2023, not peer-reviewed (cites Ruiz et al. 2017, retracted 2021)\nS4 - \"Cool pavements: a field study in Phoenix\" - Urban Climate, 2021, peer-reviewed\nS5 - \"Heat islands explained\" - Wikipedia, accessed 2026\nS6 - \"Reflective roofing and energy use\" - Energy and Buildings, 2016, peer-reviewed\nS7 - \"Urban heat and public health in Europe\" - news article, The Daily Ledger, 2024\nS8 - \"Mapping heat vulnerability with satellite data\" - Remote Sensing of Environment, 2022, peer-reviewed\n\nAvailable actions:\nsave_source(id)\nsummarize(id)\ndraft_paper(section)\nASK_HUMAN(question)\nDONE(summary)","rules":[{"type":"maxBullets","max":6},{"type":"plainText"}],"promptHints":[{"label":"Defines a good source","patterns":["peer[- ]review","2015","scholarly"]},{"label":"Flags the retraction","patterns":["retract"]},{"label":"Keeps the writing yours","patterns":["draft_paper","don'?t (write|draft)","never (write|draft)","i('ll| will) write"]},{"label":"Makes it ask before anything risky","patterns":["ask_human","ask (me|first|before)","approv","confirm","permission"]},{"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":["maxBullets","plainText"]},{"title":"Not defining a good source","why":"The AI keeps anything on topic, including Wikipedia, news and a 2012 study.","fix":"Add: \"Save only peer-reviewed sources from 2015 or later.\"","checks":["r0"]},{"title":"Not asking for red flags","why":"The AI treats a popular blog as fine without checking what it cites.","fix":"Add: \"If a source relies on a retracted study, ASK_HUMAN, don't save it.\"","checks":["r1"]},{"title":"Letting it write for you","why":"Research agents often offer to draft sections, which isn't your work.","fix":"Add: \"Never use draft_paper. End with DONE and the saved IDs.\"","checks":["r2"]}],"hint":"A research helper needs your standard for a good source. Say what qualifies, what to flag, and that the writing stays yours.","rubricLabels":["Saves S2, S4, S6 and S8, and does not save S1 (2012), S5 (Wikipedia) or S7 (news)","S3 (the blog citing a retracted study) is raised in an ASK_HUMAN line and not saved","There is no draft_paper line, and the last line is a DONE line listing the saved source IDs"]}