# Reading Plan Prompt for a Research Question > An original reusable prompt that turns one research question and your source list into a dated reading plan with time budgets and stopping rules. ## Install Copy the content below into your project: # Reading Plan Prompt for a Research Question An original reusable prompt that turns one research question and your source list into a dated reading plan with time budgets and stopping rules. ## Start here Paste your question and your source list into one ordinary AI chat that accepts text. No terminal, account setup, or browsing is needed. **Inputs to prepare** | Input | What to paste | | --- | --- | | Research question | One or two sentences | | Source list | One source per line: Title | Author or Organization | Year or date | Type | Link or where it lives | Why you saved it | | Limits | Reading time, the output you need, deadline, sources you already trust | Paste the full prompt below, fill the three input blocks, and send. Then check the reply: every source you listed appears once, minutes do not exceed your limit, each core source has an extraction block, and it ends with the closing line. If a field is missing the reply must say UNKNOWN rather than guess. **Introduction** The task is planning, not reading. The prompt works only from what you paste, so it never claims a source concludes anything and never invents titles, authors, or findings. It labels everything PLANNED. **How to check the output** Read the inventory table first: core sources should carry per-source extraction blocks. Then read the reading order against your stated time. Then read the conflict and gap map. If the total exceeds your limit, the plan must say what to drop while keeping at least one core source. **Limits and permissions** It does not browse, open links, download, or save. It cannot confirm a source exists. Empty source list: the prompt should ask you for the list instead of planning. Do not paste private or unpublished material you are not allowed to share with a chat service. Treat every claim line as your own note to fill after reading. **FAQ** **Can I use it for an undated source?** Yes. The date becomes UNKNOWN, the source is sorted as supporting, and the question is listed under Open Questions. **What if my question and my output do not match?** The first section must flag that mismatch, and the gap map names what no listed source appears to cover. **Verification note** Source reviewed; runtime not tested. Runtime verification is not performed here; this guide documents the prompt as written. **Source and thanks** Original TokRepo prompt, CC BY 4.0. Reference context: [ChatGPT release notes](), reviewed 2026-10-04. The complete prompt and its translation are appended separately. ## Complete reusable prompt # Reusable Prompt: Reproducible Reading Plan From a Question and a Source List You are a research-planning assistant. Work only from what I paste below. You cannot browse, open links, or verify that a source exists. ## My inputs 1) RESEARCH QUESTION (one or two sentences): [PASTE] 2) SOURCE LIST (one per line): Format each as: Title | Author or Organization | Year or date | Type (article, report, chapter, dataset, interview, video, web page, other) | Link or where it lives | Why I saved it (optional) [PASTE] 3) LIMITS: - Time I can spend reading: [e.g., 3 hours total, or 40 minutes per day] - Output I need: [e.g., a decision, a short essay, a study guide] - Deadline: [DATE or NONE] - Sources I already read and trust: [LIST or NONE] ## Your task Produce a reading plan that tells me exactly what to read, in what order, how long to spend, what to extract, and when to stop and reassess. Rules: - Use only the sources I listed. Never add sources, authors, dates, or findings. - If a field is missing or ambiguous, write UNKNOWN and list it in the Open Questions section. Do not guess. - Do not claim any source says anything. You have not read them. Describe what to look for, not what it concludes. - Label every recommendation as PLANNED, not as verified fact. - This plan prepares reading. It does not perform searches, downloads, or saves. ## Required output sections 1) QUESTION RESTATEMENT AND OUTPUT TARGET Restate my question in one sentence, then name the smallest useful output it must feed. Flag any mismatch between my question and my stated output. 2) SOURCE INVENTORY TABLE Columns: ID (S1, S2...), Title, Author/Org, Date, Type, For question? (core / supporting / background / unclear), Missing fields. Sort by For question? then by Date (newest first). Keep every source I listed, even weak ones, marked as such. 3) READING ORDER AND TIME BUDGET A numbered reading sequence. For each source: ID, why it appears there, planned minutes, and the single question I should answer while reading it. Fit the total into my time limit. If it does not fit, say so and propose what to drop, keeping at least one core source. 4) EXTRACTION SHEET PER SOURCE For each source, a short fill-in block: - ID and title - Claim or point (in my words after reading) - Page, section, or timestamp - Evidence strength I found: strong / weak / unclear - Supports, contradicts, or silent on my question - Follow-up needed 5) CONFLICT AND GAP MAP - Where two listed sources are likely to disagree, based on their type, date, or stated angle. - What my question needs that no listed source appears to cover. - Terms I must define before reading, drawn from my question. 6) STOPPING AND REASSESSMENT RULES Give clear conditions for: knowing enough to draft the output; needing one more source; abandoning the question as unanswerable with this list. Include a 10-minute checkpoint after each core source where I update the extraction sheets and re-decide the order. 7) OPEN QUESTIONS Every UNKNOWN field and every decision I must make before starting. End with one line: "Prepared plan only. No reading, search, or saving has occurred." ## Worked fictional example (illustration, not real data) Input: Question: "What tracking approach should our small team adopt for quarterly goals?" Sources: "Goal Tracking Case Study | J. Vale | 2023 | article | saved PDF | recommended by colleague"; "Team OKR Guide | Northwind Labs | UNKNOWN | web page | link saved"; "Quarterly Review Notes | internal | 2024 | notes | drive". Limits: 90 minutes total, output is a one-page recommendation, deadline Friday. Expected output shape: Inventory table puts the 2023 case study as core, the internal 2024 notes as core, the undated guide as supporting with Date UNKNOWN. Reading order: S1 (25 min, question: what problem did the case study solve?), S3 (20 min, question: what actually happened last quarter?), S2 (15 min, question: does the guide match either?). Conflict map notes the guide has no date, so currency is unknown. Gap map flags that no source covers team size, so I must decide whether size matters. Stopping rule: if two core sources conflict on cadence and no third source addresses it, draft the output with the conflict stated rather than resolve it. ## Checks before you return the plan - Every source I listed appears exactly once in the inventory. - No invented titles, authors, dates, or findings. - Total planned minutes do not exceed my stated limit. - Every core source has a per-source extraction block. - Conflicts are framed as likely, based on listed attributes only. - The closing line is present. ## Boundaries You are drafting a plan. You are not reading sources, searching, opening links, verifying citations, or saving anything. If my source list is empty, ask me for it instead of producing a plan. ## References and reuse - [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) · Reviewed 2026-10-04 Original TokRepo prompt · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). Reference documents retain their own rights. --- # 研究问题阅读计划提示词 一条原创可复用提示词：把研究问题和你的来源清单，变成带时间预算与停止条件的可核对阅读计划。 ## 开始使用 把问题与来源清单一并粘贴到一个能接收文本的普通 AI 对话中即可。无需终端、无需注册配置、也无需联网浏览。 **需要准备的输入** | 输入 | 粘贴内容 | | --- | --- | | 研究问题 | 一到两句话 | | 来源清单 | 每行一个来源：标题 | 作者或机构 | 年份或日期 | 类型 | 链接或存放位置 | 保存原因 | | 限制条件 | 可阅读时间、需要的产出、截止日期、你已读过并信任的来源 | 粘贴下方完整提示词，填入三个输入区块后发送。然后检查回复：你列出的每个来源只出现一次；分钟数不超过你的限制；每个核心来源都有提取块；结尾有结束语。若字段缺失，回复必须写 UNKNOWN，而不是猜测。 **介绍** 这件事是做计划，不是替你阅读。提示词只使用你粘贴的内容，因此不会声称某个来源得出了什么结论，也不会编造标题、作者或发现，一切建议都标为 PLANNED（计划）。 **如何检查输出** 先看来源清单表：核心来源应配有逐条提取块。再对照你给出的时间检查阅读顺序。然后看冲突与缺口图。若总时长超出限制，计划必须说明舍弃哪些来源，并至少保留一个核心来源。 **限制与权限** 它不浏览、不打开链接、不下载、不保存，也无法确认来源是否真实存在。来源清单为空时，它应当向你索要清单而不是直接出计划。不要粘贴你不被允许分享给对话服务的私密或未公开材料。每条“论点”都只是你阅读后自己填写的笔记。 **常见问题** **来源没有日期怎么办？** 日期写 UNKNOWN，该来源按支撑类排序，并把该问题列入开放问题。 **问题与产出不匹配怎么办？** 第一节必须指出这种不匹配，缺口图会说明没有哪个已列来源覆盖该内容。 **核对说明** 来源已审阅；运行时未测试。此处不进行运行时验证，本指南仅按提示词原文说明用法。 **来源与致谢** TokRepo 原创提示词，CC BY 4.0。参考背景：[ChatGPT release notes]()，审阅日期 2026-10-04。完整提示词及其中文翻译另行附上。 ## 完整可复制提示词 # 可复用提示词：从问题与来源清单生成可复现的阅读计划 你是一名研究规划助手。只使用我在下方粘贴的内容工作。你不能浏览、打开链接，也无法验证某个来源是否存在。 ## 我的输入 1) 研究问题（一到两句话）： [PASTE] 2) 来源清单（每行一个）： 每行格式为：标题 | 作者或机构 | 年份或日期 | 类型（文章、报告、章节、数据集、访谈、视频、网页、其他） | 链接或存放位置 | 保存原因（可选） [PASTE] 3) 限制条件： - 可投入的阅读时间：[例如：总共 3 小时，或每天 40 分钟] - 需要的产出：[例如：一个决定、一篇短文、一份学习指南] - 截止日期：[DATE 或 NONE] - 我已读过并信任的来源：[LIST 或 NONE] ## 你的任务 产出一份阅读计划，明确告诉我该读什么、按什么顺序读、每项花多长时间、要提取什么，以及何时停下来重新评估。 规则： - 只使用我列出的来源。绝不添加来源、作者、日期或发现。 - 若某字段缺失或含糊，写 UNKNOWN，并列入开放问题部分。不要猜测。 - 不要声称任何来源说了什么。你没有读过它们。描述该去找什么，而不是它得出了什么结论。 - 把每条建议标为 PLANNED，而不是已验证的事实。 - 本计划是为阅读做准备。它不执行搜索、下载或保存。 ## 必需的输出部分 1) 问题复述与产出目标 用一句话复述我的问题，然后说出它必须服务的最小可用产出。指出我的问题与我所述产出之间是否存在任何不匹配。 2) 来源清单表 列：ID（S1、S2……）、标题、作者/机构、日期、类型、对问题的作用？（核心 / 支撑 / 背景 / 不清楚）、缺失字段。 按 对问题的作用？ 排序，再按 日期 排序（最新在前）。保留我列出的每个来源，即使很弱，也要如实标注。 3) 阅读顺序与时间预算 一个编号的阅读序列。对每个来源：ID、它为何出现在该位置、计划分钟数，以及我在阅读时应当回答的单个问题。让总时长符合我的时间限制。若放不下，就说明这一点，并提议舍弃哪些来源，同时至少保留一个核心来源。 4) 每个来源的提取表 对每个来源，一个简短的填空块： - ID 与标题 - 论点或观点（阅读后用我自己的话写） - 页码、章节或时间戳 - 我发现的证据强度：强 / 弱 / 不清楚 - 对问题的支持、矛盾或未涉及 - 需要的后续工作 5) 冲突与缺口图 - 根据两个已列来源的类型、日期或所述角度，它们可能在何处不一致。 - 我的问题需要但没有哪个已列来源似乎覆盖的内容。 - 我必须在阅读前定义的术语，取自我的问题。 6) 停止与重新评估规则 给出明确的判断条件，用于确定：已有足够信息来起草产出；还需要再找一个来源；用这份清单问题无法回答，应予放弃。在每读完一个核心来源后安排一个 10 分钟检查点，用来更新提取表并重新决定顺序。 7) 开放问题 每个 UNKNOWN 字段，以及我在开始前必须做出的每个决定。 结尾写一行：“Prepared plan only. No reading, search, or saving has occurred.” ## 虚构的示例演示（仅作说明，非真实数据） 输入：问题：“我们小团队为季度目标应采用什么跟踪方法？”来源：“Goal Tracking Case Study | J. Vale | 2023 | article | saved PDF | recommended by colleague”；“Team OKR Guide | Northwind Labs | UNKNOWN | web page | link saved”；“Quarterly Review Notes | internal | 2024 | notes | drive”。限制条件：总共 90 分钟，产出是一页建议，截止日期为周五。 预期输出形态：来源清单表把 2023 年案例研究列为核心里程，把 2024 年内部笔记列为核心，把无日期指南列为支撑来源，日期为 UNKNOWN。阅读顺序：S1（25 分钟，问题：该案例研究解决了什么问题？），S3（20 分钟，问题：上一季度实际发生了什么？），S2（15 分钟，问题：该指南与上述两者是否吻合？）。冲突图指出该指南没有日期，因此时效性未知。缺口图标明没有来源覆盖团队规模，所以我必须决定规模是否重要。停止规则：如果两个核心来源在节奏上冲突，且没有第三个来源涉及该问题，就带着已说明的冲突起草产出，而不是强行解决它。 ## 返回计划前的检查 - 我列出的每个来源在清单中恰好出现一次。 - 没有编造的标题、作者、日期或发现。 - 计划的总分钟数不超过我说明的限制。 - 每个核心来源都有一个逐来源提取块。 - 冲突仅基于已列属性，被表述为可能存在的冲突。 - 结束语已出现。 ## 边界 你在起草一份计划。你不在阅读来源、搜索、打开链接、核实引文或保存任何内容。如果我的来源清单为空，请向我索要清单，而不是生成计划。 ## 参考资料与复用 - [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) · Reviewed 2026-10-04 TokRepo 原创提示词 · [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)。参考资料保留各自原有权利。 --- Source: https://tokrepo.com/en/workflows/reading-plan-prompt-research-question-b1135a21 Author: Prompt Lab