deep-reading-coach
Use this skill whenever the user wants to read, understand, summarize, analyze, evaluate, compare, or build notes from a book, PDF, article, chapter, or multiple sources. Trigger for precise or vague English/Chinese reading-understanding requests, including "help me read this", "what is this about", "summarize this", "digest this", "is this worth reading", "make a reading plan", "help me truly understand this", "compare these books", "study this topic", "帮我看看", "这讲啥", "帮我读这本书", "总结一下", "整理重点", "消化一下", "这本书值不值得读", "帮我真正读懂", "比较这些书", or "围绕这个主题做研究". Do not trigger for purely mechanical PDF/file operations such as extracting one page's text, OCR only, converting file formats, searching keywords, exporting tables, counting words, or listing images unless the user also asks for understanding, reading guidance, summary, analysis, evaluation, or synthesis. Infer the likely reading goal, choose the right reading mode automatically, hide formal reading-method jargon from ordinary users, and guide the user through active reading rather than merely summarizing for them.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/fanpuyen--deep-reading-coach)HTML
<a href="/listings/fanpuyen--deep-reading-coach"><img src="/api/listings/fanpuyen--deep-reading-coach/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
用户级 Codex skill 通常放在:
%USERPROFILE%\.codex\skills\deep-reading-coach
安装或修改后,请重启 Codex,让新线程加载这个 skill。
推荐的 GitHub 仓库结构:
skills/
deep-reading-coach/
SKILL.md
README.md
README_EN.md
references/
routing.md
templates.md
pdf-intake.md
scripts/
pdf_intake.py从 GitHub 安装或手动复制时,请安装/复制整个 `deep-reading-coach` 文件夹,而不是只复制 `SKILL.md`。`references` 和 `scripts` 都是这个 skill 的一部分。
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
SKILL.md · Manifest · 177 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- ---
- name: deep-reading-coach
- description: Use this skill whenever the user wants to read, understand, summarize, analyze, evaluate, compare, or build notes from a book, PDF, article, chapter, or multiple sources. Trigger for precise or vague English/Chinese reading-understanding requests, including "help me read this", "what is this about", "summarize this", "digest this", "is this worth reading", "make a reading plan", "help me truly understand this", "compare these books", "study this topic", "帮我看看", "这讲啥", "帮我读这本书", "总结一下", "整理重点", "消化一下", "这本书值不值得读", "帮我真正读懂", "比较这些书", or "围绕这个主题做研究". Do not trigger for purely mechanical PDF/file operations such as extracting one page's text, OCR only, converting file formats, searching keywords, exporting tables, counting words, or listing images unless the user also asks for understanding, reading guidance, summary, analysis, evaluation, or synthesis. Infer the likely reading goal, choose the right reading mode automatically, hide formal reading-method jargon from ordinary users, and guide the user through active reading rather than merely summarizing for them.
- ---
- # Deep Reading Coach
- Help the user read actively and think with a text. The method is inspired by *How to Read a Book*, but ordinary users should see plain labels: **Quick Preview**, **Deep Breakdown**, **Topic Study**, and **Reading Plan**.
- Do not reproduce long copyrighted passages. Paraphrase methods and cite page numbers, chapter names, headings, or short compliant excerpts when helpful.
- ## Reference Loading
- Keep this file as the dispatcher. Load reference files only when needed:
- - `references/routing.md`: vague intent, near-miss cases, mode selection, summary handling, and book-type request routing.
- - `references/templates.md`: full bilingual output templates, plus optional compact templates for explicitly brief requests.
- - `references/analytical-outline.md`: analytical structural outline rules for Rule 3 when the user asks for analytical reading, whole-book structure, the author's problem, argument movement, or serious book notes.
- - `references/book-types.md`: book-type adapters for practical books, literature, history, science, mathematics, philosophy, social science, and related forms.
- - `references/pdf-intake.md`: manual PDF reliability protocol when the script cannot run.
- - `references/ocr-routing.md`: OCR route selection for scanned PDFs, including Windows built-in OCR, AI vision OCR preparation, and Tesseract/Poppler fallback.
- Use bundled scripts when available:
- - `scripts/pdf_intake.py`: diagnose PDF readability, extract page-level text, and produce `intake_report.json`, `intake_report.md`, and `pages.jsonl`.
- - `scripts/windows_ocr_pdf.py`: use Windows built-in OCR on embedded PDF page images; preferred on Windows for scanned Chinese PDFs when available.
- - `scripts/prepare_vision_ocr.py`: export PDF page images, manifest, batches, and a prompt for AI vision OCR.
- - `scripts/ocr_pages.py`: run page-level OCR for scanned PDFs when local `pdftoppm` and `tesseract` tools are available; otherwise write a clear missing-tools diagnostic.
- - `scripts/check_report.py`: check generated reading reports for required structural sections and guardrails.
- ## Language Policy
- - Respond in the user's primary language.
- - If the source language differs from the user's language, explain in the user's language while preserving important original terms when useful.
- - Use bilingual labels only when useful for method explanation or glossary work.
- ## When Not To Use This Skill
- This skill is for reading comprehension, reading guidance, structured notes, analysis, evaluation, and synthesis.
- Do not use it for purely mechanical file/PDF operations such as OCR only, extracting one page's text, converting file formats, searching keywords, exporting tables, counting words/pages/images, listing metadata, or translating a specific passage without reading analysis.
- Use it when a file operation is paired with comprehension, such as "OCR this PDF and then help me understand it" or "extract the text and summarize the argument."
- ## First Move
- 1. Detect language.
- 2. Infer mode and depth from the user's wording and context.
- 3. If a PDF is provided, run `scripts/pdf_intake.py` before reading when local file access is available. If it cannot run, use `references/pdf-intake.md`.
- - If the PDF is text-based, read the extracted page text with page references.
- - If the PDF is scanned, image-based, mixed, or low quality, do not stop at "OCR recommended." Load `references/ocr-routing.md` and actively try the available OCR route before making page-level claims.
- - On Windows, try `scripts/windows_ocr_pdf.py` first. If Windows OCR is unavailable or too weak, prepare AI vision OCR inputs with `scripts/prepare_vision_ocr.py`. Use `scripts/ocr_pages.py` only when Poppler/pdftoppm and Tesseract are installed.
- 4. If a non-PDF file is provided, inspect structure before making content claims.
- 5. If no source is available, say so and work from the user's description or excerpts.
- 6. State the working assumption briefly when intent is vague, then proceed.
- ## Mode Routing
- Use `references/routing.md` for detailed examples. Default assumptions:
- | User intent | Default mode |
- |---|---|
- | "Help me look/read this", "帮我看看" | Quick Preview |
- | "What is this about?", "这讲啥" | Lightweight summary + Quick Preview |
- | "Summarize this", "总结一下" | Concise summary; offer Quick Preview or Deep Breakdown if useful |
- | "Is this worth reading?", "该怎么读" | Quick Preview / Reading Plan |
- | "I want to truly understand", "深度拆解" | Full Deep Breakdown / Analytical Reading |
- | "Do analytical reading", "分析阅读", "15 条规则" | Full 15-rule Analytical Reading |
- | "Compare/study this topic" | Topic Study |
- Low confidence: ask one short question. Do not ask users to choose between formal method names unless they already used them.
- ## Output Defaults
- - **Quick Preview default**: use the full **Inspectional Reading Report / 检视阅读报告** from `references/templates.md`.
- - Use **Compact Quick Preview** only when the user explicitly asks for a brief/short/compressed overview, e.g. "简单说", "简短版", "只要大概", "brief preview", or "short version".
- - **Deep Breakdown default**: use the full 15-rule Analytical Reading template from `references/templates.md`.
- - Use **Compact Deep Breakdown** only when the user explicitly asks for a brief/short/compressed deep reading.
- - If the user asks for a narrow part, answer that part directly, but keep the analysis-reading boundaries intact.
- All reading modes above elementary reading should address the four basic active-reading questions. Quick Preview answers them provisionally; Deep Breakdown answers them formally; Topic Study answers them across sources.
- ## Topic Study Core
- Topic Study corresponds to syntopical reading. Use it when the user studies one question across multiple books, articles, chapters, or excerpts.
- Do not write separate summaries and then compare them loosely. Let the user's question govern the reading, and make sources serve that question.
- Topic Study has two stages:
- 1. Preparation: build a trial bibliography or source pool, inspect sources quickly, decide which sources are relevant, and revise the research question if needed.
- 2. Syntopical reading: locate relevant passages, build a neutral vocabulary, clarify shared questions, define issues from different answers, and arrange the discussion objectively.
- Use the full Topic Study template in `references/templates.md` by default for multi-source/topic-comparison requests. Mark direct answers, indirect answers, and silence separately. Do not adopt one author's terminology as the master vocabulary.
- ## Quick Preview Boundary
- Quick Preview is inspectional reading: orientation, provisional judgment, reading decision, and route planning. It is not analytical reading.
- It may say what the text appears to be, how it seems organized, what may deserve close reading, and what questions should move into Deep Breakdown.
- It must not formally define key concepts, establish shared terms, extract detailed propositions, reconstruct arguments, evaluate truth, or produce a full analytical outline. Those belong to Deep Breakdown.
- ## Deep Breakdown Core
- Deep Breakdown corresponds to analytical reading. It may make formal judgments after working through the text carefully.
- In Stage 1 of Deep Breakdown, do not merely restate the Quick Preview. Build structural understanding: classify with evidence, write a testable whole-book statement, map the book's skeleton, and turn the author's purpose into a question tree.
- For Rule 3, load `references/analytical-outline.md` when the user asks for analytical reading, whole-book structure, the author's problem, argument movement, or serious book notes. The outline must be an analytical structural outline, not a copied table of contents or chapter-by-chapter summary. It should identify the governing question, functional major parts, transitions between parts, chapter functions, and the reasoning used to construct the outline.
- In Stage 2, interpret the book by reading unit before synthesizing the whole. Track chapter/section-level key terms, important sentences, propositions, and local arguments, then connect them into a whole-book argument map and return to the Stage 1 question tree.
- In Stage 3, do not write a generic pros/cons review. First prove understanding, then judge by scope: agree, partly agree, disagree, or suspend judgment. Tie criticism to specific propositions, argument IDs, or unanswered questions, and separate knowledge-based judgment from personal preference.
- Deep Breakdown should cover:
- 1. What is the text about as a whole?
- 2. What exactly is being said, and how?
- 3. Is it true, sound, or persuasive?
- 4. What follows for the user?
- Full Analytical Reading follows these 15 rules:
- 1. Classify the book by kind and subject.
- 2. State what the whole book is about as briefly as possible.
- 3. Outline the whole and its major parts in order and relation.
- 4. Find the author's question or problem.
- 5. Interpret key words and reach shared meanings.
- 6. Grasp the author's propositions from important sentences.
- 7. Reconstruct the author's arguments.
- 8. Determine what is solved, unsolved, or knowingly unsolved.
- 9. Do not criticize until you can say you understand.
- 10. Do not disagree in a contentious spirit.
- 11. Give reasons for criticism and distinguish knowledge from opinion.
- 12. Show where the author is uninformed.
- 13. Show where the author is misinformed.
- 14. Show where the author is illogical.
- 15. Show where the author's analysis or account is incomplete.
- After Rule 1, use `references/book-types.md` when the book type should change how Rules 2-15 are applied. The adapter adjusts the reading; it does not replace the rules.
- ## PDF Intake
- When a PDF is provided, do not jump directly into reading claims. First determine whether it is text-based, scanned, mixed, page-image-heavy, or unreliable.
- Use `scripts/pdf_intake.py` and reflect its result in the output. Preserve page references when reading extracted text. State limitations when extraction quality is medium or low.
- For scanned or low-quality PDFs, load `references/ocr-routing.md`. Keep the analysis honest without becoming shallow. Distinguish confirmed evidence, reasoned inference, and claims needing OCR or rereading. If an OCR workflow or OCR-capable tool is available, use it before making page-level claims; otherwise say that OCR is needed and produce only a provisional structural reading.
- OCR route preference for Windows:
- 1. Use `scripts/windows_ocr_pdf.py` when Windows OCR is available.
- 2. Use `scripts/prepare_vision_ocr.py` when a vision-capable model or OCR service should read page images.
- 3. Use `scripts/ocr_pages.py` only when Poppler/pdftoppm and Tesseract are installed.
- ## Long-Text Pattern
- For long books, stage the work:
- 1. Quick Preview.
- 2. User chooses chapters or questions.
- 3. Deep Breakdown on selected sections.
- 4. Evaluation and personal application.
- 5. Optional Topic Study with related sources.
- ## Quality Checklist
- Before finalizing, verify:
- - The output language follows the user.
- - The mode and depth match the prompt.
- - Full reading reports are used by default unless the user asks for a brief version.
- - PDF/source limitations are stated.
- - Understanding comes before judgment.
- - Author claims are separated from assistant interpretation.
- - The user's next action is clear.
README.md · README · 362 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- # Deep Reading Coach
- 中文 | [English](README_EN.md)
- `deep-reading-coach` 是一个 Codex skill,用来帮助你更认真、更有方法地读书、读 PDF、读文章,或者围绕一个主题比较多份资料。
- 它不是普通的“总结器”,更像一个阅读教练。它会先帮你判断资料怎么读,再帮你拆结构、抓概念、看论证,最后把阅读结果和你的目标联系起来。
- ---
- ## 它能做什么
- 适合用来:
- - 快速判断一本书值不值得读
- - 给一本书或 PDF 制定阅读路线
- - 提炼一本书的主题、结构、重点章节
- - 深度拆解作者的问题、概念、主张和论证
- - 生成结构化读书笔记
- - 围绕一个主题比较多本书或多篇文章
- - 把阅读结果转成问题清单、洞察或行动建议
- ---
- ## 快速开始
- 在 Codex 里直接说:
- ```text
- 使用 deep-reading-coach 帮我读这本书。
- ```
- 如果你有文件路径,可以这样说:
- ```text
- 使用 deep-reading-coach 阅读 <PDF_PATH>,先帮我看看这本书该怎么读。
- ```
- 你不需要知道“检视阅读、分析阅读、主题阅读”这些术语。这个 skill 会根据你的说法自动选择合适的方式。
- ---
- ## 模糊说法也可以
- 你不需要把指令写得很标准。很口语的说法也可以:
- ```text
- 帮我看看这个。
- ```
- ```text
- 这讲啥?
- ```
- ```text
- 帮我消化一下。
- ```
- ```text
- 我想搞懂这本书,但不知道从哪里开始。
- ```
- 这个 skill 会自动判断应该先做快速预览、简明总结、深度拆解、主题研究,还是阅读计划。如果实在判断不出来,它应该只问一个简短问题,而不是让你选择专业术语。
- ---
- ## 什么时候不该用它
- 这个 skill 适合“读懂、总结、分析、评价、比较、做笔记”。如果你只是做机械文件处理,不一定需要它。
- 不适合只用它处理:
- - 提取某一页文字
- - 只做 OCR
- - 把 PDF 转成 Word/Markdown
- - 搜索某个关键词
- - 导出表格到 Excel
- - 统计字数、页数、图片数
- - 列出 PDF 里的图片或元数据
- - 只翻译某一页或某一段
- 如果你同时要求理解,就适合用它:
- ```text
- 把这个 PDF OCR 后帮我读懂。
- ```
- ```text
- 提取文字并整理核心观点。
- ```
- ```text
- 把这一章转成读书笔记。
- ```
- ---
- ## PDF 读取
- 当你提供 PDF 时,这个 skill 应该先检查 PDF 是否能可靠读取,而不是直接开始分析。
- 它自带一个辅助脚本:
- ```text
- scripts/pdf_intake.py
- ```
- 这个脚本会诊断:
- - 页数
- - 是否能直接提取文字
- - 是否像扫描版或图片版 PDF
- - 空页或低质量页面
- - 是否有乱码/异常提取
- - 是否识别到目录
- - 哪些页面需要复查
- - 建议读取方式
- 典型命令:
- ```text
- python scripts/pdf_intake.py path/to/file.pdf --out work/pdf-intake
- ```
- 如果 PDF 是文本型,skill 会保留页码并读取提取文本。如果 PDF 是扫描版或混合型,skill 会先标明证据等级:哪些是已确认,哪些是根据目录和结构做出的合理推断,哪些需要 OCR 或重读确认,而不是假装已经完整读懂。
- 在 Windows 上,优先使用系统内置 OCR,不需要 Poppler/Tesseract:
- ```text
- python scripts/windows_ocr_pdf.py path/to/file.pdf --out work/windows-ocr --lang zh-Hans-CN
- ```
- 它会尝试直接从 PDF 抽取页面图片,再调用 Windows OCR 识别中文,并生成逐页文本。
- 如果你想使用 AI 视觉模型识别页面图片,可以先导出视觉 OCR 输入包:
- ```text
- python scripts/prepare_vision_ocr.py path/to/file.pdf --out work/vision-ocr-input
- ```
- 它会导出页面图片、manifest、批处理 JSON 和 `vision_ocr_prompt.md`。这条路线适合 Windows OCR 效果差、旧书扫描复杂,或者你准备使用支持视觉的模型时。
- 如果本机有 `pdftoppm` 和 `tesseract`,也可以用 Tesseract fallback:
- ```text
- python scripts/ocr_pages.py path/to/file.pdf --out work/pdf-ocr
- ```
- 它会生成 `ocr_report.md`、`ocr_pages.jsonl` 和逐页 OCR 文本。如果缺少 OCR 工具,它会写出缺失工具诊断,不会偷偷安装任何东西。
- ---
- ## 常见用法
- ### 1. Quick Preview / 检视阅读
- 适合你想先知道一本书讲什么、值不值得读、该怎么读。
- Quick Preview / 检视阅读是一轮快速定向,帮助你判断这本书表面上讲什么、值不值得读、该怎么读。它对应《如何阅读一本书》里的检视阅读,包括两个动作:**有系统的略读或粗读** 和 **粗浅的第一遍阅读**。
- 它不是完整分析阅读,不负责正式解释关键概念、主旨和论证。它只做暂定观察和阅读决策:初步判断书的类型、表层结构、哪些部分值得读、哪些问题要带入下一轮。同时,它也必须暂定回答四个基本问题:整体在谈什么、细部说了什么和怎么说、是否初步看来有道理、跟读者有什么关系。如果你想真正读懂一本书,应继续进入 Deep Breakdown / 分析阅读。
- ```text
- 使用 deep-reading-coach 帮我快速看一下这本 PDF,告诉我它讲什么、适合谁、哪些章节值得细读。
- ```
- ```text
- 这本书我还没决定要不要读,先帮我判断值不值得读。
- ```
- 默认会输出完整检视阅读报告。只有你明确要求“简短版”“压缩版”“只要大概”时,才会使用短版。
- 通常会输出:
- - 来源/PDF 读取情况
- - 目录是否可见
- - 阅读决策:跳过、略读、选读、精读,还是作为参考资料
- - 暂定书籍类型,以及为什么影响读法
- - 这本书大概像是在讲什么
- - 表层结构地图
- - 四个基本问题的暂定回答
- - 有系统略读记录
- - 粗浅第一遍阅读记录
- - 针对不同目标的阅读路线
- - 哪些部分应精读、略读或暂时跳过
- - 带入分析阅读的问题
- - 是否进入分析阅读、阅读计划、主题研究,或暂停
- ### 2. Deep Breakdown / 分析阅读
- 适合你想真正读懂一本书,而不只是看摘要。这个模式对应《如何阅读一本书》里的分析阅读,会按 15 条规则分三阶段处理。第一阶段会特别重视“透视一本书”:用证据分类、写出可检验的全书大意、拆出分析阅读式全书大纲,并整理作者的问题树。
- 分析阅读式全书大纲不是目录摘要。它要找出作者的统摄性问题、全书推进公式、功能性大部分、各章在论证中的作用,以及这些判断是怎样从书名、副标题、目录、章节标题、序言、结尾或正文证据中推出来的。
- 第二阶段会按章节或关键部分诠释内容:先建立阅读单元地图,再分别找出全书关键词、局部关键词、关键句、主旨、局部论述链,最后合成为全书论述地图,并回到第一阶段的问题树判断作者解决了什么。
- 第三阶段会做公正评论:先证明已经理解作者,再按范围判断同意、部分同意、不同意或暂缓评论。批评必须绑定具体主旨、论述或未回答的问题,并区分知识判断和个人偏好。
- ```text
- 使用 deep-reading-coach 帮我深度拆解这本书,重点看作者想解决什么问题、核心概念、主要主张和论证结构。
- ```
- ```text
- 我想认真读懂这本书,不只是总结。请帮我做一份深度阅读笔记。
- ```
- 默认会输出完整分析阅读 / 深度拆解。只有你明确要求“简短版”“压缩版”“先给我大概”时,才会使用短版。
- 通常会输出:
- - 这本书整体在谈什么
- - 正式分类、分类证据、读法后果与书类适配器
- - 一句话全书大意,以及对象、核心问题、主要回答、展开方式
- - 分析阅读式全书大纲:统摄性问题、全书推进公式、功能性大部分、各章功能表、各部分之间的转接关系
- - 大纲构造说明:哪些来自文本证据,哪些是合理推断,哪些需要 OCR、重读或查证
- - 作者想解决的问题树:主问题、次问题、先后顺序、对应章节
- - 阅读单元地图:哪些章节 / 部分被精读,哪些暂缓
- - 全书核心关键字、局部关键字与共通词义
- - 按章节 / 部分找出的关键句与主旨 / 命题
- - 局部论述链、论述之间的关系、全书总论述地图
- - 回到问题树判断作者解决了什么,部分解决了什么,没解决什么
- - 评论资格检查:先确认理解,再进行同意、不同意或暂缓评论
- - 作者观点的最强版本,以及我的判断是否需要调整
- - 四种严格批评:作者知识不足、知识错误、不合逻辑、分析不完整
- - “这本书说得有道理吗”的综合判断
- - 这本书对你的理解、判断、行动、研究和个人目标有什么关系
- - 这本书的使用边界和误用风险
- 如果你想检查生成报告是否遗漏结构,可以运行:
- ```text
- python scripts/check_report.py path/to/report.md --type analytical
- ```
- 不同种类的书不会被用同一把尺子硬套。例如:
- - 实用书重点看目标、手段、是否可执行,以及读者同意后要采取什么行动。
- - 文学作品重点看整体经验、人物、情节、意象、形式,不会强行把小说或诗当成论证文章。
- - 历史书重点看史料、选择、解释、因果关系,以及是否需要比较多本历史书。
- - 科学和数学书重点看定义、假设、实验、证明、图表、公式和证据范围。
- - 哲学书重点看问题、关键概念、区分、前提、推理和结论。
- - 社会科学书重点看术语定义、方法、数据、价值立场,并经常建议进入主题研究。
- ### 3. 主题研究
- 适合你围绕一个主题比较多本书、多篇文章、多个章节或多段资料。它不是把资料分别摘要后并排比较,而是让“你要研究的问题”成为中心,所有资料都服务于这个问题。
- 主题研究通常分两步:
- 1. 准备阶段:先建立试验性书目或资料池,快速检视资料,判断哪些真正相关,并在必要时修订研究问题。
- 2. 正式主题阅读:找出相关章节 / 段落,建立中立词汇,厘清共同问题,从不同回答中界定议题,再客观排列整个讨论。
- ```text
- 使用 deep-reading-coach 帮我围绕“学习方法”做主题研究,列出应该比较的问题和书单。
- ```
- ```text
- 我有三本关于投资的书,帮我比较它们对风险、复利和资产配置的看法。
- ```
- 通常会输出:
- - 起始研究问题,以及检视后是否需要修订
- - 试验性书目 / 资料池
- - 检视筛选表:纳入、排除、暂缓及理由
- - 每份资料真正相关的章节、段落、页码或论述
- - 中立词汇表:避免直接采用某一个作者的术语
- - 共同问题表:哪些问题是多份资料都能回答的
- - 不同资料对同一问题的直接回答、间接回答或沉默
- - 议题地图:主要分歧在哪里、各自站在哪一边
- - 四个基本问题的跨资料比较
- - 共识、分歧与沉默
- - 辩证客观检查:是否公平排列讨论,没有过早站队
- - 综合判断:你现在可以合理得出什么结论
- - 还没解决的问题
- ### 4. 阅读计划
- 适合你不知道该怎么读,或者时间有限。
- ```text
- 我不知道怎么读这本书,使用 deep-reading-coach 帮我制定一个一周阅读计划。
- ```
- ```text
- 我只有两个小时,帮我安排这本书的阅读顺序。
- ```
- 通常会输出:
- - 阅读目标
- - 时间分配
- - 先读哪些章节
- - 哪些内容可以跳过或略读
- - 读书笔记模板
- - 后续深读建议
- ---
- ## 推荐工作流
- 最舒服的用法通常是:
- 1. 先让它做快速预览。
- 2. 你挑出最重要的章节或问题。
- 3. 再让它做深度拆解。
- 4. 如果需要,再扩展到相关书籍做主题研究。
- 例如:
- ```text
- 使用 deep-reading-coach 先快速预览这本 PDF,告诉我哪些章节最值得深读。
- ```
- 然后继续说:
- ```text
- 好,接下来深度拆解第 2、3、5 章。
- ```
- ---
- ## 不适合做什么
- 这个 skill 不适合用来:
- - 大段复制受版权保护的书籍内容
- - 在没有资料的情况下假装读过某本书
- - 替代你自己的判断
- - 对明显只需要略读的资料强行写长篇分析
- 如果你只是想要非常短的摘要,也可以用它;但它可能会提醒你,摘要只是第一步,真正读懂还需要看结构、概念、论证和与你自己的关系。
- ---
- ## 安装
- 用户级 Codex skill 通常放在:
- ```text
- %USERPROFILE%\.codex\skills\deep-reading-coach
- ```
- 安装或修改后,请重启 Codex,让新线程加载这个 skill。
- 推荐的 GitHub 仓库结构:
- ```text
- skills/
- deep-reading-coach/
- SKILL.md
- README.md
- README_EN.md
- references/
- routing.md
- templates.md
- pdf-intake.md
- scripts/
- pdf_intake.py
- ```
- 从 GitHub 安装或手动复制时,请安装/复制整个 `deep-reading-coach` 文件夹,而不是只复制 `SKILL.md`。`references` 和 `scripts` 都是这个 skill 的一部分。
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.