# DYZ MedPaper — full text > 32 个医学论文写作与科研 Agent Skills / 32 medical writing and research Agent Skills. 仓库 / Repository: https://github.com/duyunzhoudyz/dyz-med-paper-skills · 文档站 / Docs: https://duyunzhoudyz.github.io/dyz-med-paper-skills/ ## 常见问题 / FAQ **DYZ MedPaper 是什么?** DYZ MedPaper 是一套面向中国医生与医学研究生的医学论文写作与科研 Agent Skills 集合,包含 32 个技能,覆盖选题、PubMed 文献检索、统计分析与结果解读、中英写作、选刊投稿与审稿修回。源码仓库为 github.com/duyunzhoudyz/dyz-med-paper-skills,文档站为 duyunzhoudyz.github.io/dyz-med-paper-skills。当前为公开测试版,采用 PolyForm Noncommercial 1.0.0 许可。 **可以在哪些智能体里使用?** 任何支持 Agent Skills 的宿主都可以加载完整技能目录:Claude Code 安装到 ~/.claude/skills,Codex 默认安装到 ~/.agents/skills,也可用 --target 指定其他宿主的技能目录。技能由 Markdown 说明、模板、参考资料和本地 Python 脚本组成,不绑定某个模型厂商;实际的任务路由与触发效果需要在目标宿主中核实。 **32 个技能分别覆盖什么?** 技能按五组组织:起点与项目状态 5 个、研究设计与文献证据 7 个、统计分析与图表 5 个、论证与论文写作 7 个、翻译修订与交付 8 个,合计 32 个。完整清单见 docs/skills.md;skills.json 还给出每个技能的中英标题、输入输出、触发词和指令地址。 **如何安装并开始第一项任务?** 支持三种零依赖安装方式:1)双击 installers/install-windows.bat 或 installers/install-macos.command;2)Claude Code 2.1.275+ 可用 /plugin install dyz-med-all --marketplace duyunzhoudyz/dyz-med-paper-skills 或按 5 个阶段轻量包安装(dyz-med-start/evidence/analysis/writing/delivery);3)终端运行 python scripts/install_codex.py [--pack all|start|evidence|analysis|writing|delivery] [--target 目录] 并加 --check 校验。需要 Python 3.10+。 **是否免费?许可如何?** 源码公开,采用 PolyForm Noncommercial License 1.0.0:许可证允许的非商业用途可以免费使用;超出许可范围需要单独书面商业授权,见仓库中的 COMMERCIAL-LICENSE.md。第三方材料遵循各自条款。 **能自动产出一篇可直接投稿的论文吗?** 不能。技能提供流程、模板和可追溯的检查点;正式统计分析、数据库访问、复杂排版和最终医学判断仍需要人和相应软件环境。内置在线检索脚本只覆盖 PubMed,知网、万方、维普等需要合法访问权限。教学示例均为合成材料,不是已发表研究或使用效果证明。 **What is DYZ MedPaper?** DYZ MedPaper is a set of 32 medical writing and research Agent Skills for doctors and medical postgraduate students in China. It covers research questions, PubMed literature search, statistical analysis and result interpretation, Chinese and English writing, journal selection and reviewer responses. The source repository is github.com/duyunzhoudyz/dyz-med-paper-skills and the documentation site is duyunzhoudyz.github.io/dyz-med-paper-skills. It is a public beta under the PolyForm Noncommercial 1.0.0 license. **Which agents can use it?** Any host that supports Agent Skills can load the complete skill directories: Claude Code installs into ~/.claude/skills, Codex installs into ~/.agents/skills by default, and --target points at another host's skill directory. Each skill is Markdown instructions plus templates, references and local Python scripts, with no dependency on one model vendor. Actual task routing and triggering still need to be verified in the target host. **What do the 32 skills cover?** The skills are organised in five groups: 5 starting points and project state, 7 study design and literature evidence, 5 statistical analysis and figures, 7 argument and manuscript writing, and 8 translation, revision and delivery, 32 in total. The complete list is in docs/skills.md; skills.json adds each skill's bilingual title, inputs, outputs, trigger text and instructions URL. **How do I install it and start the first task?** Three stdlib-only options are available: (1) double-click installers/install-windows.bat or installers/install-macos.command; (2) in Claude Code 2.1.275+, install the full suite or one of 5 stage packs via /plugin install dyz-med-all --marketplace duyunzhoudyz/dyz-med-paper-skills; or (3) run python scripts/install_codex.py [--pack all|start|evidence|analysis|writing|delivery] [--target DIR] followed by --check. Requires Python 3.10+. **Is it free, and under which license?** The source is public under the PolyForm Noncommercial License 1.0.0: noncommercial use within the license is free, and use beyond it needs a separate written commercial license, described in COMMERCIAL-LICENSE.md. Third-party materials keep their own terms. **Does it produce a submission-ready paper automatically?** No. The skills provide workflows, templates and traceable checkpoints; formal statistical analysis, database access, advanced typesetting and final medical judgement still need people and the relevant software. The built-in online search script covers PubMed only; CNKI, Wanfang and VIP need legitimate access. Teaching examples are synthetic and are not published studies or evidence of results. ## 技能完整清单 / Complete skill catalog ## dyz-med-ideate — 医学科研选题 从临床观察梳理可验证的研究问题 - 输入 / Input: 临床观察、兴趣与资源限制 - 输出 / Output: 候选问题与可行性路线 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-ideate/SKILL.md ## dyz-med-init — 论文项目初始化 为已明确的研究建立论文项目与结构 - 输入 / Input: 研究问题、设计与已有材料 - 输出 / Output: 项目骨架与材料缺口 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-init/SKILL.md ## dyz-med-manuscript-diagnose — 医学稿件诊断 先评估已有稿件的问题与修改顺序 - 输入 / Input: 初稿、摘要或拒稿意见 - 输出 / Output: 诊断报告与修改优先级 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-manuscript-diagnose/SKILL.md ## dyz-med-orchestrator — 论文进度与下一步 根据现有材料判断下一步任务 - 输入 / Input: 论文文件夹与当前需求 - 输出 / Output: 进度地图与下一步建议 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-orchestrator/SKILL.md ## dyz-med-passport — 项目状态记录 内部维护已确认事实与产出版本 - 输入 / Input: 已确认事实、材料和批准记录 - 输出 / Output: 可追溯的项目状态 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-passport/SKILL.md ## dyz-med-mechanism — 机制与因果逻辑 检查机制假设、因果路径与证据层级 - 输入 / Input: 研究问题、设计与支持证据 - 输出 / Output: 机制链、假设缺口与验证路径 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-mechanism/SKILL.md ## dyz-med-data-adequacy — 数据充分性检查 判断数据能支持的问题和分析范围 - 输入 / Input: 数据字典、设计与获准数据 - 输出 / Output: 充分性报告与可分析范围 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-data-adequacy/SKILL.md ## dyz-med-registration — 研究注册核查 对照官方注册记录与方案、稿件 - 输入 / Input: 注册号、方案与稿件 - 输出 / Output: 注册对照表与待核实差异 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-registration/SKILL.md ## dyz-med-cnki — PubMed 与中文文献检索 设计 PubMed 和中文医学来源检索并记录覆盖范围 - 输入 / Input: 研究问题、数据库与纳排范围 - 输出 / Output: 检索式、记录与候选文献池 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-cnki/SKILL.md ## dyz-med-key-literature — 关键文献精读 对已有文献作结构化阅读与证据比较 - 输入 / Input: 候选文献与可访问全文 - 输出 / Output: 单篇简报与证据地图 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-key-literature/SKILL.md ## dyz-med-screening-pipeline — 系统综述文献筛选 管理去重、筛选、纳排理由与 PRISMA 计数 - 输入 / Input: 检索记录、筛选协议与全文 - 输出 / Output: 筛选记录、提取表与计数 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-screening-pipeline/SKILL.md ## dyz-med-evidence-audit — 论断与引用核查 检查文献和数据是否支持稿件论断 - 输入 / Input: 稿件、来源文献与研究结果 - 输出 / Output: 论断—证据矩阵与问题定位 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-evidence-audit/SKILL.md ## dyz-med-analyze — 医学统计分析 先形成计划,再按批准范围执行统计 - 输入 / Input: 获准数据、数据字典与研究问题 - 输出 / Output: 分析计划、脚本与结果记录 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-analyze/SKILL.md ## dyz-med-interpret — 统计结果与临床意义 区分统计证据、效应大小与临床重要性 - 输入 / Input: 统计结果与有来源的临床阈值 - 输出 / Output: 逐结局解释与谨慎措辞 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-interpret/SKILL.md ## dyz-med-sensitivity — 敏感性与稳健性分析 围绕明确的不确定性比较替代假设 - 输入 / Input: 主要结果、具体假设与批准范围 - 输出 / Output: 稳健性比较与局限说明 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-sensitivity/SKILL.md ## dyz-med-figures-tables — 医学论文图表 规划和核对 Table 1、森林图等图表 - 输入 / Input: 已执行分析输出与数据来源 - 输出 / Output: 图表、图注与一致性清单 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-figures-tables/SKILL.md ## dyz-med-flowchart — 研究流程图 由真实计数整理 PRISMA、CONSORT 等流程图 - 输入 / Input: 研究流程、筛选记录与计数 - 输出 / Output: 流程图、图注与计数核对 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-flowchart/SKILL.md ## dyz-med-writing-plan — 论文论证与写作提纲 把研究问题、结果与证据组织成段落计划 - 输入 / Input: 研究问题、已核实结果与文献 - 输出 / Output: 论证地图与章节提纲 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-writing-plan/SKILL.md ## dyz-med-introduction — 医学论文引言 围绕具体知识缺口组织引言 - 输入 / Input: 研究问题、写作提纲与核实文献 - 输出 / Output: 引言草稿与逻辑说明 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-introduction/SKILL.md ## dyz-med-methods — 医学论文方法部分 将实际研究过程写成可复核方法 - 输入 / Input: 方案、执行记录与分析计划 - 输出 / Output: 方法正文与缺项清单 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-methods/SKILL.md ## dyz-med-results — 医学论文结果部分 把核实后的统计输出写成结果正文 - 输入 / Input: 已核实统计输出、表格和图 - 输出 / Output: 结果正文与数字对应关系 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-results/SKILL.md ## dyz-med-discussion — 医学论文讨论 组织主要发现、证据比较、意义与局限 - 输入 / Input: 已核实结果、研究设计与文献 - 输出 / Output: 讨论草稿与论断边界 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-discussion/SKILL.md ## dyz-med-abstract — 标题、摘要与关键词 从已完成正文生成结构化摘要 - 输入 / Input: 完整正文、核实结果与期刊要求 - 输出 / Output: 标题、摘要与关键词 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-abstract/SKILL.md ## dyz-med-write — 全文初稿与统稿 按明确请求整合材料或统一已有章节 - 输入 / Input: 已确认研究材料或完成的章节 - 输出 / Output: 完整初稿或统稿版本 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-write/SKILL.md ## dyz-med-translate — 医学论文中英翻译 翻译正文与回复并保护数字、引用和术语 - 输入 / Input: 原文、术语表与目标语言 - 输出 / Output: 译文、术语表与保护检查 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-translate/SKILL.md ## dyz-med-deai — 学术语言自然化 改善重复空泛表达并保留事实和作者语气 - 输入 / Input: 原稿与表达偏好 - 输出 / Output: 自然化稿件与修改说明 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-deai/SKILL.md ## dyz-med-guideline — 医学研究报告规范 核对 STROBE、CONSORT、PRISMA 等适用条目 - 输入 / Input: 稿件、研究类型与适用规范 - 输出 / Output: 条目覆盖表与定位建议 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-guideline/SKILL.md ## dyz-med-review — 投稿前模拟审稿 从证据、方法等视角评估稿件 - 输入 / Input: 接近完成的稿件与附件 - 输出 / Output: 模拟审稿报告与修改建议 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-review/SKILL.md ## dyz-med-revision — 改稿与审稿意见回复 逐条落实修改并记录正文位置 - 输入 / Input: 稿件、审稿意见与已完成改动 - 输出 / Output: 修订稿、逐条回复与修改定位 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-revision/SKILL.md ## dyz-med-journal-finder — 医学论文选刊 按主题、读者、预算与当前要求筛选期刊 - 输入 / Input: 稿件主题、文章类型、预算与时间 - 输出 / Output: 分层候选期刊与核实依据 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-journal-finder/SKILL.md ## dyz-med-export — Word 审阅稿导出 整理章节、引用和图表生成基础 DOCX - 输入 / Input: 完成的章节、图表与文献 - 输出 / Output: 基础 DOCX 审阅稿与导出报告 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-export/SKILL.md ## dyz-med-submit — 投稿材料准备 准备投稿信、题名页与声明等本地文件 - 输入 / Input: 稿件、作者确认信息与期刊要求 - 输出 / Output: 投稿包与待确认清单 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-submit/SKILL.md ## dyz-med-ideate — Research questions Use for clinical research ideation, PICO questions and feasibility exploration. - 输入 / Input: Clinical observations, interests and constraints - 输出 / Output: Candidate questions and feasibility routes - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-ideate/SKILL.md ## dyz-med-init — Project setup Use to initialize a medical manuscript project from an established study. - 输入 / Input: Research question, design and available materials - 输出 / Output: Project structure and material gaps - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-init/SKILL.md ## dyz-med-manuscript-diagnose — Manuscript diagnosis Use for read-only manuscript assessment and revision priorities. - 输入 / Input: Draft, abstract or rejection feedback - 输出 / Output: Diagnosis and prioritized next steps - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-manuscript-diagnose/SKILL.md ## dyz-med-orchestrator — Workflow navigation Use to review medical research workflow progress and recommend the next task. - 输入 / Input: Paper workspace and current request - 输出 / Output: Progress map and next-step recommendation - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-orchestrator/SKILL.md ## dyz-med-passport — Project state Use internally to maintain confirmed project facts, provenance and artifact versions. - 输入 / Input: Confirmed facts, artifacts and approvals - 输出 / Output: Traceable project state - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-passport/SKILL.md ## dyz-med-mechanism — Mechanisms and causality Use to examine causal reasoning, mechanism hypotheses and study logic. - 输入 / Input: Research question, design and evidence - 输出 / Output: Mechanism chain, gaps and validation plan - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-mechanism/SKILL.md ## dyz-med-data-adequacy — Data adequacy Use to assess medical research data adequacy before statistical analysis. - 输入 / Input: Data dictionary, design and authorized data - 输出 / Output: Adequacy report and defensible analysis scope - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-data-adequacy/SKILL.md ## dyz-med-registration — Study registration Use to check trial registration records against protocols and manuscripts. - 输入 / Input: Registry identifier, protocol and manuscript - 输出 / Output: Registration comparison and unresolved differences - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-registration/SKILL.md ## dyz-med-cnki — Medical literature search Use for PubMed and Chinese medical literature search; database execution requires access. - 输入 / Input: Question, databases and eligibility scope - 输出 / Output: Queries, search log and candidate references - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-cnki/SKILL.md ## dyz-med-key-literature — Critical literature reading Use for critical appraisal and synthesis of available full-text medical literature. - 输入 / Input: Candidate papers and accessible full texts - 输出 / Output: Paper briefs and evidence map - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-key-literature/SKILL.md ## dyz-med-screening-pipeline — Systematic review screening Use for systematic review and meta-analysis screening, extraction and PRISMA counts. - 输入 / Input: Search records, screening protocol and full texts - 输出 / Output: Screening log, extraction tables and counts - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-screening-pipeline/SKILL.md ## dyz-med-evidence-audit — Claim–evidence audit Use to audit manuscript claims against cited literature and study evidence. - 输入 / Input: Manuscript, cited sources and study results - 输出 / Output: Claim–evidence matrix and located issues - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-evidence-audit/SKILL.md ## dyz-med-analyze — Biostatistical analysis Use for biostatistical analysis planning and approved R or other local statistical execution. - 输入 / Input: Authorized data, dictionary and research question - 输出 / Output: Analysis plan, scripts and recorded results - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-analyze/SKILL.md ## dyz-med-interpret — Clinical interpretation Use to interpret existing statistical results and clinical importance without rerunning models. - 输入 / Input: Existing results and sourced clinical thresholds - 输出 / Output: Outcome-level interpretation and calibrated wording - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-interpret/SKILL.md ## dyz-med-sensitivity — Sensitivity analysis Use for approved sensitivity analyses and robustness comparisons of existing findings. - 输入 / Input: Primary results, assumptions and approved scope - 输出 / Output: Robustness comparisons and limitations - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-sensitivity/SKILL.md ## dyz-med-figures-tables — Manuscript tables and figures Use for medical manuscript tables, statistical figures and text–figure consistency. - 输入 / Input: Executed analysis outputs and provenance - 输出 / Output: Figures, tables, captions and consistency checklist - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-figures-tables/SKILL.md ## dyz-med-flowchart — Study flow diagrams Use for PRISMA, CONSORT and observational study flow diagrams from traceable counts. - 输入 / Input: Study flow, screening records and counts - 输出 / Output: Flow diagram, caption and count checks - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-flowchart/SKILL.md ## dyz-med-writing-plan — Manuscript writing plan Use for medical manuscript argument maps and paragraph-level writing plans. - 输入 / Input: Question, verified results and literature - 输出 / Output: Argument map and section outline - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-writing-plan/SKILL.md ## dyz-med-introduction — Introduction writing Use to draft or revise a medical manuscript introduction and research rationale. - 输入 / Input: Question, outline and verified literature - 输出 / Output: Introduction draft and rationale - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-introduction/SKILL.md ## dyz-med-methods — Methods writing Use to write methods sections from actual study procedures and approved analysis plans. - 输入 / Input: Protocol, execution records and analysis plan - 输出 / Output: Methods draft and missing-information list - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-methods/SKILL.md ## dyz-med-results — Results writing Use to write results sections while preserving verified effect estimates and numerical provenance. - 输入 / Input: Verified statistical outputs, tables and figures - 输出 / Output: Results draft and numerical traceability - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-results/SKILL.md ## dyz-med-discussion — Discussion writing Use to draft or revise medical discussions, evidence comparisons and limitations. - 输入 / Input: Verified results, design and literature - 输出 / Output: Discussion draft and limits of claims - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-discussion/SKILL.md ## dyz-med-abstract — Titles and abstracts Use for medical manuscript titles, structured abstracts and bilingual abstracts. - 输入 / Input: Completed manuscript, verified results and journal rules - 输出 / Output: Title, abstract and keywords - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-abstract/SKILL.md ## dyz-med-write — Full manuscript drafting Use for explicitly requested full manuscript drafts or coherence editing across completed sections. - 输入 / Input: Confirmed research materials or completed sections - 输出 / Output: Full draft or coordinated revision - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-write/SKILL.md ## dyz-med-translate — Medical translation Use for Chinese–English medical translation with numerical, citation and terminology preservation. - 输入 / Input: Source text, glossary and target language - 输出 / Output: Translation, glossary and protected-content checks - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-translate/SKILL.md ## dyz-med-deai — Academic style editing Use to improve repetitive academic prose while preserving facts and disclosure; not detector evasion. - 输入 / Input: Draft and style preferences - 输出 / Output: Edited prose and change notes - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-deai/SKILL.md ## dyz-med-guideline — Reporting guidelines Use to check applicable medical reporting guidelines including STROBE, CONSORT and PRISMA. - 输入 / Input: Manuscript, study type and applicable guideline - 输出 / Output: Coverage checklist and located recommendations - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-guideline/SKILL.md ## dyz-med-review — Pre-submission review Use for simulated pre-submission peer review and located revision recommendations. - 输入 / Input: Near-complete manuscript and supporting files - 输出 / Output: Simulated review and revision recommendations - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-review/SKILL.md ## dyz-med-revision — Revision and reviewer responses Use for manuscript revision, response to reviewers and journal resubmission preparation. - 输入 / Input: Manuscript, reviewer comments and completed changes - 输出 / Output: Revision, point-by-point response and change locations - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-revision/SKILL.md ## dyz-med-journal-finder — Journal selection Use for medical journal selection based on verified scope, requirements and fees. - 输入 / Input: Topic, article type, budget and timeline - 输出 / Output: Tiered journal shortlist and verified rationale - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-journal-finder/SKILL.md ## dyz-med-export — Word review draft export Use to assemble basic DOCX review drafts; advanced formatting requires suitable external tools. - 输入 / Input: Completed sections, figures, tables and references - 输出 / Output: Basic DOCX review draft and export report - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-export/SKILL.md ## dyz-med-submit — Submission package Use to prepare local cover letters and journal submission packages; external submission needs authorization. - 输入 / Input: Manuscript, confirmed author details and journal rules - 输出 / Output: Submission package and confirmation checklist - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-submit/SKILL.md ## 页面正文 / Page text ### Project Status, Validation and License https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/about/ ## Current status DYZ MedPaper is a public beta collection of medical manuscript Agent Skills for doctors and medical postgraduate students in China. It includes 32 skills, templates, references, synthetic examples and local scripts. | Check | What it establishes | What it does not establish | |---|---|---| | Installation and file checks | Complete directory copies and byte consistency | Correct loading or invocation in a host | | Offline regression and synthetic workflows | Explicitly implemented tool and artifact behavior | Model writing quality or medical validity | | Documentation build checks | Page, link, language and indexing structure | Search-engine indexing or higher rankings | Real-host routing, sustained writing and independent medical quality have not all completed acceptance. See [live offline checks](https://github.com/duyunzhoudyz/dyz-med-paper-skills/actions/workflows/check.yml), [validation details](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/docs/readme-reference.md#validation) and [release notes](https://github.com/duyunzhoudyz/dyz-med-paper-skills/releases). ## Tools and access - The built-in online search script covers PubMed (reference identity checks also surface Crossref/PubMed retraction, expression-of-concern and erratum alerts). CNKI, Wanfang, VIP and other databases require lawful access. Discovery leads do not establish a completed search or full-text retrieval. - Formal statistical execution needs approval of a specific plan, R or another established engine, and the necessary packages; stdlib helpers cover local PHI screening, Table 1, KM survival, binary DL meta-analysis, diagnostic 2x2/ROC metrics and provenance-bound SVG figures. - Local exporters produce basic DOCX review drafts (`export_docx.py`) and native Word tracked-changes drafts (`export_tracked_docx.py` with ``). Advanced layout, equations and visual inspection still require Word or LibreOffice. - Numerical, reporting-checklist (6 full-subitem official checklists) and citation checks cover implemented deterministic comparisons; they do not certify semantic support or formal compliance. ## License The project is source-available under [PolyForm Noncommercial License 1.0.0](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/LICENSE). Uses outside the license's permitted noncommercial scope require [separate written commercial authorization](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/COMMERCIAL-LICENSE.md). Third-party materials retain their own terms. ## Feedback and contributions The maintainer, public source and change history are available on the [GitHub project page](https://github.com/duyunzhoudyz/dyz-med-paper-skills). The "Current status" and "Tools and access" sections above describe which checks run in local scripts and which still depend on the model and human review. Report your system, host, reproduction steps and de-identified errors in [GitHub Issues](https://github.com/duyunzhoudyz/dyz-med-paper-skills/issues). Improvements to examples, terminology and installation documentation are welcome. Do not upload patient records or credentials. [Contributing](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/CONTRIBUTING.md) · [Maintaining the site and releases](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/docs/maintaining-discoverability.md) · [Generative engine optimization](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/docs/geo-playbook.md) · [Project descriptions and links](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/docs/discovery-kit.md) ### Medical Writing & Research Agent Skills https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/ [Install and get started →](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/install/){: .primary-link} [Explore all 32 skills](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/skills/){: .secondary-link} **Work in Chinese. Write in Chinese or English.** Plan a study question, search medical literature, structure a manuscript and turn supervisor or reviewer feedback into specific revisions. Start with one task or keep working around the same paper. ## Start with the materials you already have | Your starting point | What to work on | Task guide | |---|---|---| | A clinical question | Search scope, existing evidence and unanswered questions | [PubMed and Chinese literature search](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/pubmed/) | | Results and verified papers | An argument map and IMRAD sections | [Medical manuscript writing](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/writing/) | | Statistical outputs | Effect size, uncertainty and clinical importance | [Interpreting statistical results](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/statistics/) | | Review search records | Screening decisions and PRISMA counts | [Systematic review workflow](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/systematic-review/) | | A Chinese manuscript | Translation with numerical and citation checks | [Medical translation](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/translation/) | | Reviewer feedback | Point-by-point responses and located changes | [Responding to reviewers](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/reviewer-response/) | ## Built around clinical research in China Discuss questions and revisions in Chinese before preparing Chinese or English prose. The workflow can account for departmental case data, follow-up constraints, supervisor feedback, graduation requirements, journal scope and publication budgets. Most skill instructions and supporting materials are in Chinese; this site provides an English entry point. You do not need to memorize skill names. A useful first request is: > Here are my results table, verified references and discussion draft. Map the main findings, then revise the discussion. Separate evidence from interpretation, list unsupported claims and explain the changes in Chinese. An existing draft can go straight to diagnosis and revision. Reviewer feedback can go straight to a response plan. One invocation does not automatically complete a paper. ## Keep the files and the reasoning Each skill bundles a `SKILL.md` file with relevant templates, references and local scripts. Store research inputs and outputs in a separate paper workspace. Outlines, drafts, literature records and revision logs remain available for supervisors and collaborators. The built-in exporter produces basic Word review drafts. [GitHub repository](https://github.com/duyunzhoudyz/dyz-med-paper-skills) · [Status and license](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/about/) · [Releases](https://github.com/duyunzhoudyz/dyz-med-paper-skills/releases) ## Current scope This is a public beta. Database searches require access, and the built-in online search script covers PubMed only. Statistical execution needs an appropriate environment and approval of the specific plan. Real-host behavior, sustained writing and independent medical quality have not all been validated. Teaching examples are synthetic, not publication outcomes or user testimonials. **What is DYZ MedPaper?** DYZ MedPaper is a set of 32 medical writing and research Agent Skills for doctors and medical postgraduate students in China. It covers research questions, PubMed literature search, statistical analysis and result interpretation, Chinese and English writing, journal selection and reviewer responses. The source repository is github.com/duyunzhoudyz/dyz-med-paper-skills and the documentation site is duyunzhoudyz.github.io/dyz-med-paper-skills. It is a public beta under the PolyForm Noncommercial 1.0.0 license. **Which agents can use it?** Any host that supports Agent Skills can load the complete skill directories: Claude Code installs into ~/.claude/skills, Codex installs into ~/.agents/skills by default, and --target points at another host's skill directory. Each skill is Markdown instructions plus templates, references and local Python scripts, with no dependency on one model vendor. Actual task routing and triggering still need to be verified in the target host. **What do the 32 skills cover?** The skills are organised in five groups: 5 starting points and project state, 7 study design and literature evidence, 5 statistical analysis and figures, 7 argument and manuscript writing, and 8 translation, revision and delivery, 32 in total. The complete list is in docs/skills.md; skills.json adds each skill's bilingual title, inputs, outputs, trigger text and instructions URL. **How do I install it and start the first task?** Three stdlib-only options are available: (1) double-click installers/install-windows.bat or installers/install-macos.command; (2) in Claude Code 2.1.275+, install the full suite or one of 5 stage packs via /plugin install dyz-med-all --marketplace duyunzhoudyz/dyz-med-paper-skills; or (3) run python scripts/install_codex.py [--pack all|start|evidence|analysis|writing|delivery] [--target DIR] followed by --check. Requires Python 3.10+. **Is it free, and under which license?** The source is public under the PolyForm Noncommercial License 1.0.0: noncommercial use within the license is free, and use beyond it needs a separate written commercial license, described in COMMERCIAL-LICENSE.md. Third-party materials keep their own terms. **Does it produce a submission-ready paper automatically?** No. The skills provide workflows, templates and traceable checkpoints; formal statistical analysis, database access, advanced typesetting and final medical judgement still need people and the relevant software. The built-in online search script covers PubMed only; CNKI, Wanfang and VIP need legitimate access. Teaching examples are synthetic and are not published studies or evidence of results. ### Install Medical Writing Skills in Claude Code and Codex https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/install/ ## 1. Prepare a host and materials Use an account-configured agent that supports Agent Skills and Python 3.10+. The installer uses the standard library; it does not install models, statistical packages or database accounts. Create a separate paper workspace containing synthetic or authorized, de-identified materials. Local script execution does not make the whole AI session offline: the model service may receive conversation content. ## 2. Download the repository Choose **Code → Download ZIP** on [GitHub](https://github.com/duyunzhoudyz/dyz-med-paper-skills), extract it and open a terminal in the directory, or use Git: ```bash git clone https://github.com/duyunzhoudyz/dyz-med-paper-skills.git cd dyz-med-paper-skills ``` On Windows, check `python --version`; substitute `py -3` if that is your Python launcher. On macOS/Linux, `python3` is often appropriate. ## 3. Install for your host ### Claude Code ```bash python scripts/install_codex.py --target "~/.claude/skills" python scripts/install_codex.py --target "~/.claude/skills" --check ``` ### Codex ```bash python scripts/install_codex.py python scripts/install_codex.py --check ``` The default destination is `~/.agents/skills/`. The filename remains `install_codex.py`; `--target` supports other hosts' documented skill directories. Success reports `PASS`, the skill count and destination. Different existing content stops installation instead of silently replacing edits. ### Install for one paper project Replace the example with an existing project directory. For Claude Code: ```bash python scripts/install_codex.py --target "/path/to/my-paper/.claude/skills" ``` For Codex: ```bash python scripts/install_codex.py --project "/path/to/my-paper" ``` Keep the same target arguments and add `--check` to verify. Choose personal or project scope for each host to avoid duplicate entries. ## 4. Start a task Open the paper workspace in your host and reload its skills. Describe the task naturally: > Use dyz-med-orchestrator to inspect the available manuscript materials, identify the most useful next step and help with that task. List missing inputs separately. Explicit invocation is `/dyz-med-orchestrator` in Claude Code or `$dyz-med-orchestrator` in Codex. File checks do not certify that a host has loaded or routed a skill correctly; verify that in the host. ## Upgrade and troubleshoot Run `--check` with the original destination arguments first. To upgrade from a new checkout, provide a new backup directory outside both the source and installation directories: ```bash python scripts/install_codex.py --upgrade --backup "/path/to/new-backup" ``` For Claude Code, also keep `--target "~/.claude/skills"`. Manual installation must include complete skill directories and the repository's `LICENSE`, `NOTICE` and `COMMERCIAL-LICENSE.md`, not just `SKILL.md` files. [Full maintenance instructions](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/README_EN.md#maintenance) · [Skill catalog](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/skills/) · [Task guides](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/) ### 32 Medical Writing and Research Agent Skills https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/skills/ Choose the task that matches your materials; completing every stage is optional. Examples are synthetic teaching materials. Most full skill instructions and supporting resources are in Chinese, with bilingual task descriptions for discovery. ### dyz-med-ideate — Research questions Use for clinical research ideation, PICO questions and feasibility exploration. - 输入 / Input: Clinical observations, interests and constraints - 输出 / Output: Candidate questions and feasibility routes - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-ideate/SKILL.md ### dyz-med-init — Project setup Use to initialize a medical manuscript project from an established study. - 输入 / Input: Research question, design and available materials - 输出 / Output: Project structure and material gaps - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-init/SKILL.md ### dyz-med-manuscript-diagnose — Manuscript diagnosis Use for read-only manuscript assessment and revision priorities. - 输入 / Input: Draft, abstract or rejection feedback - 输出 / Output: Diagnosis and prioritized next steps - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-manuscript-diagnose/SKILL.md ### dyz-med-orchestrator — Workflow navigation Use to review medical research workflow progress and recommend the next task. - 输入 / Input: Paper workspace and current request - 输出 / Output: Progress map and next-step recommendation - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-orchestrator/SKILL.md ### dyz-med-passport — Project state Use internally to maintain confirmed project facts, provenance and artifact versions. - 输入 / Input: Confirmed facts, artifacts and approvals - 输出 / Output: Traceable project state - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-passport/SKILL.md ### dyz-med-mechanism — Mechanisms and causality Use to examine causal reasoning, mechanism hypotheses and study logic. - 输入 / Input: Research question, design and evidence - 输出 / Output: Mechanism chain, gaps and validation plan - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-mechanism/SKILL.md ### dyz-med-data-adequacy — Data adequacy Use to assess medical research data adequacy before statistical analysis. - 输入 / Input: Data dictionary, design and authorized data - 输出 / Output: Adequacy report and defensible analysis scope - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-data-adequacy/SKILL.md ### dyz-med-registration — Study registration Use to check trial registration records against protocols and manuscripts. - 输入 / Input: Registry identifier, protocol and manuscript - 输出 / Output: Registration comparison and unresolved differences - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-registration/SKILL.md ### dyz-med-cnki — Medical literature search Use for PubMed and Chinese medical literature search; database execution requires access. - 输入 / Input: Question, databases and eligibility scope - 输出 / Output: Queries, search log and candidate references - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-cnki/SKILL.md ### dyz-med-key-literature — Critical literature reading Use for critical appraisal and synthesis of available full-text medical literature. - 输入 / Input: Candidate papers and accessible full texts - 输出 / Output: Paper briefs and evidence map - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-key-literature/SKILL.md ### dyz-med-screening-pipeline — Systematic review screening Use for systematic review and meta-analysis screening, extraction and PRISMA counts. - 输入 / Input: Search records, screening protocol and full texts - 输出 / Output: Screening log, extraction tables and counts - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-screening-pipeline/SKILL.md ### dyz-med-evidence-audit — Claim–evidence audit Use to audit manuscript claims against cited literature and study evidence. - 输入 / Input: Manuscript, cited sources and study results - 输出 / Output: Claim–evidence matrix and located issues - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-evidence-audit/SKILL.md ### dyz-med-analyze — Biostatistical analysis Use for biostatistical analysis planning and approved R or other local statistical execution. - 输入 / Input: Authorized data, dictionary and research question - 输出 / Output: Analysis plan, scripts and recorded results - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-analyze/SKILL.md ### dyz-med-interpret — Clinical interpretation Use to interpret existing statistical results and clinical importance without rerunning models. - 输入 / Input: Existing results and sourced clinical thresholds - 输出 / Output: Outcome-level interpretation and calibrated wording - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-interpret/SKILL.md ### dyz-med-sensitivity — Sensitivity analysis Use for approved sensitivity analyses and robustness comparisons of existing findings. - 输入 / Input: Primary results, assumptions and approved scope - 输出 / Output: Robustness comparisons and limitations - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-sensitivity/SKILL.md ### dyz-med-figures-tables — Manuscript tables and figures Use for medical manuscript tables, statistical figures and text–figure consistency. - 输入 / Input: Executed analysis outputs and provenance - 输出 / Output: Figures, tables, captions and consistency checklist - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-figures-tables/SKILL.md ### dyz-med-flowchart — Study flow diagrams Use for PRISMA, CONSORT and observational study flow diagrams from traceable counts. - 输入 / Input: Study flow, screening records and counts - 输出 / Output: Flow diagram, caption and count checks - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-flowchart/SKILL.md ### dyz-med-writing-plan — Manuscript writing plan Use for medical manuscript argument maps and paragraph-level writing plans. - 输入 / Input: Question, verified results and literature - 输出 / Output: Argument map and section outline - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-writing-plan/SKILL.md ### dyz-med-introduction — Introduction writing Use to draft or revise a medical manuscript introduction and research rationale. - 输入 / Input: Question, outline and verified literature - 输出 / Output: Introduction draft and rationale - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-introduction/SKILL.md ### dyz-med-methods — Methods writing Use to write methods sections from actual study procedures and approved analysis plans. - 输入 / Input: Protocol, execution records and analysis plan - 输出 / Output: Methods draft and missing-information list - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-methods/SKILL.md ### dyz-med-results — Results writing Use to write results sections while preserving verified effect estimates and numerical provenance. - 输入 / Input: Verified statistical outputs, tables and figures - 输出 / Output: Results draft and numerical traceability - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-results/SKILL.md ### dyz-med-discussion — Discussion writing Use to draft or revise medical discussions, evidence comparisons and limitations. - 输入 / Input: Verified results, design and literature - 输出 / Output: Discussion draft and limits of claims - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-discussion/SKILL.md ### dyz-med-abstract — Titles and abstracts Use for medical manuscript titles, structured abstracts and bilingual abstracts. - 输入 / Input: Completed manuscript, verified results and journal rules - 输出 / Output: Title, abstract and keywords - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-abstract/SKILL.md ### dyz-med-write — Full manuscript drafting Use for explicitly requested full manuscript drafts or coherence editing across completed sections. - 输入 / Input: Confirmed research materials or completed sections - 输出 / Output: Full draft or coordinated revision - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-write/SKILL.md ### dyz-med-translate — Medical translation Use for Chinese–English medical translation with numerical, citation and terminology preservation. - 输入 / Input: Source text, glossary and target language - 输出 / Output: Translation, glossary and protected-content checks - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-translate/SKILL.md ### dyz-med-deai — Academic style editing Use to improve repetitive academic prose while preserving facts and disclosure; not detector evasion. - 输入 / Input: Draft and style preferences - 输出 / Output: Edited prose and change notes - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-deai/SKILL.md ### dyz-med-guideline — Reporting guidelines Use to check applicable medical reporting guidelines including STROBE, CONSORT and PRISMA. - 输入 / Input: Manuscript, study type and applicable guideline - 输出 / Output: Coverage checklist and located recommendations - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-guideline/SKILL.md ### dyz-med-review — Pre-submission review Use for simulated pre-submission peer review and located revision recommendations. - 输入 / Input: Near-complete manuscript and supporting files - 输出 / Output: Simulated review and revision recommendations - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-review/SKILL.md ### dyz-med-revision — Revision and reviewer responses Use for manuscript revision, response to reviewers and journal resubmission preparation. - 输入 / Input: Manuscript, reviewer comments and completed changes - 输出 / Output: Revision, point-by-point response and change locations - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-revision/SKILL.md ### dyz-med-journal-finder — Journal selection Use for medical journal selection based on verified scope, requirements and fees. - 输入 / Input: Topic, article type, budget and timeline - 输出 / Output: Tiered journal shortlist and verified rationale - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-journal-finder/SKILL.md ### dyz-med-export — Word review draft export Use to assemble basic DOCX review drafts; advanced formatting requires suitable external tools. - 输入 / Input: Completed sections, figures, tables and references - 输出 / Output: Basic DOCX review draft and export report - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-export/SKILL.md ### dyz-med-submit — Submission package Use to prepare local cover letters and journal submission packages; external submission needs authorization. - 输入 / Input: Manuscript, confirmed author details and journal rules - 输出 / Output: Submission package and confirmation checklist - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-submit/SKILL.md Database searches, statistical execution and advanced formatting need the relevant access, environment and tools. `passport` maintains state internally; use `orchestrator` to review project progress. [Install complete skill directories](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/install/) · [Capabilities and validation scope](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/about/) ### PubMed and Chinese Medical Literature Search https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/pubmed/ `dyz-med-cnki` is the literature-search skill's directory name. Its scope includes PubMed and Chinese medical sources. The built-in online script covers PubMed; CNKI, Wanfang, VIP and other sources require suitable tools and access. ## Define the question and scope Provide the population, exposure or intervention, comparator, outcomes and study designs, together with date, language and database limits. A review project also needs a search protocol and eligibility criteria. > Design a PubMed and Chinese-database search for this question. Identify core concepts, synonyms and subject headings to verify. Record actual access for each source and list sources that could not be searched. ## Execute and preserve the record 1. Agree on concepts and queries, then execute in accessible databases. 2. Save database names, search dates, actual queries, applied limits and result counts. 3. Export native records with provenance. Local import tools support RIS, literal BibTeX, NBIB and PubMed XML. 4. Preserve merge relationships during deduplication. Keep candidate discovery separate from formal systematic-review screening. Web leads without database execution remain `discovery_only`. Abstract-only access does not count as full-text critical reading. ## Move from candidates to evidence Use `dyz-med-key-literature` to read accessible full texts in a structured way: design, population, effect estimates, limitations and relevance to the question. Include contradictory evidence and applicability differences alongside supporting studies. > Select key studies from the verified candidate pool. Distinguish guidelines, reviews, primary studies and contradictory evidence. State which sources were read in full and which are abstract-only. Outputs include a search plan, execution log, reference pool, deduplication record and paper briefs. See the synthetic [search example](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-cnki/examples/worked-example.md) and [local literature tool instructions](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-cnki/references/literature-cli.md), both in Chinese. For reviews, continue to [screening and PRISMA](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/systematic-review/). With reliable evidence ready, proceed to [manuscript writing](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/writing/). ### Response to Reviewers for Medical Papers https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/reviewer-response/ Use `dyz-med-revision` to revise a manuscript, prepare responses or adapt it for another journal. If a rejection letter leaves the main problem unclear, start with a read-only `manuscript-diagnose` assessment. ## Gather the inputs Provide the current manuscript, editor letter, complete comments separated by reviewer, journal requirements and changes already made. Keep version identifiers and page or line references for reliable locations. > Number the comments and build a revision plan. Separate wording changes from facts needing confirmation and requests requiring reanalysis. Describe a change as completed only after it has actually been made. ## Build a response matrix | Field | Content | |---|---| | Identifier and original comment | One-to-one correspondence with the decision letter | | Issue and action | Explanation, text change, analysis or supported disagreement | | Manuscript change | Specific modification and location | | Draft response | Answer, rationale and pointer to the revision | | Status | Completed, awaiting confirmation or pending execution | Do not turn a planned analysis into a claim that it was performed. Confirm new data, reanalysis, ethics, authorship or other source facts before recording actual results. ## Write responses that point to changes Answer the substantive question and identify the change and its location. If disagreeing, explain the reason in relation to the design and evidence, and consider clarifying the manuscript. Courtesy alone does not answer the comment. > Update the responses using the completed manuscript changes. Include a location for each change and retain unfinished items as pending. Compare response commitments against the manuscript and list what remains. ## Check and deliver Deliver the revised manuscript, response letter, change log and open-task list. Reconcile the abstract, results, discussion and figures. `submit` prepares local submission materials; external upload or final submission requires explicit authorization and final confirmation. The synthetic [reviewer-response example](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-revision/examples/worked-example-reviewer-response.md) demonstrates locations and pending work. [Revision and delivery skills](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/skills/#delivery) · [Medical translation](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/translation/) ### Medical Statistics and Clinical Importance https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/statistics/ The statistical tasks have distinct owners: `data-adequacy` assesses what data can support; `analyze` plans and executes approved analyses; `interpret` explains existing results; and `sensitivity` examines specific changes in assumptions. ## Choose a starting point | What you have | Appropriate task | |---|---| | Raw data only | Inspect structure, missingness and adequacy | | A need for primary analysis | Prepare a reviewable plan before execution | | Existing outputs | Interpret effects, intervals and clinical importance | | Concern about an assumption | Define the uncertainty and plan sensitivity work | ## Interpret existing results Provide effect scales, units, reference groups, confidence intervals, populations and model details. Clinical-importance discussion also needs prespecified, sourced thresholds; record missing thresholds rather than inventing them. > Interpret these existing results without refitting models. For each outcome, explain effect size and interval uncertainty, separating statistical evidence from clinical importance. State where a clinical threshold has not been established. Check direction, units and reference groups. A nonsignificant finding is not automatically evidence of no effect, and a P value alone does not replace interpretation of effect estimates and uncertainty. ## If additional analysis is needed Identify the concrete question, such as a missing-data assumption or analysis-population change. Specify alternative assumptions, required inputs, outputs and execution conditions in a plan, then work within the approved scope. Preserve scripts, logs, diagnostics and result versions. ## Hand off to writing Deliver an outcome-level interpretation table, threshold questions, calibrated wording and proposed additional analyses. `results` writes from verified numbers; `discussion` considers design and evidence; `figures-tables` checks consistency across displays and prose. [Interpretation example](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-interpret/examples/worked-example.md) · [Analysis example](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-analyze/examples/worked-example-cohort.md) · [Manuscript writing](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/writing/) These synthetic examples are teaching materials. Researchers and statistical professionals must review outputs; tool results do not certify study conclusions. ### Systematic Review Screening and PRISMA https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/systematic-review/ Start with an established search protocol and database records. `screening-pipeline` maintains the screening evidence trail, `flowchart` prepares diagrams, and `guideline` checks applicable reporting items. ## Required inputs Provide the protocol, eligibility criteria, native database exports, accessible full texts and screening arrangements. Identify sources not covered. A PubMed-only search does not establish completed multi-database coverage. > Import these records with source and deduplication provenance. Build title/abstract and full-text screening tables from the protocol. Preserve independent decisions, disagreement resolution and full-text exclusion reasons. List missing inputs first. ## Keep three levels distinct | Level | Record | |---|---| | Search record | Source query, date, returned count and duplicates | | Report | Citation, full-text access, eligibility decision and exclusion reason | | Study | Related reports, extraction and risk-of-bias assessment | Preserve independent screening judgments and how disagreements were resolved. Multiple reports may describe one study; paper counts and study counts are not interchangeable. Keep missing full texts and pending decisions visible. ## Build and reconcile the flow diagram Check stage totals and branches before asking `flowchart` to prepare a PRISMA diagram and caption. Count validation does not replace eligibility judgment. > Derive PRISMA counts from completed screening records, separating records, reports and studies. Identify unfinished screening or retrieval steps rather than reporting them as completed. ## Before a meta-analysis Clarify extraction, report-to-study relationships and the appropriate bias assessment before choosing a synthesis model or sensitivity analysis. `analyze` handles statistical work only within an approved plan and available environment. Installing the skills does not install statistical software or packages. [Screening example](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-screening-pipeline/examples/worked-example.md) · [Flow-diagram example](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-flowchart/examples/worked-example.md) · [Literature-search guide](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/pubmed/) These are synthetic teaching examples in Chinese, not records of a completed real review. ### Chinese–English Medical Translation https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/translation/ Use `dyz-med-translate` for Chinese-to-English, English-to-Chinese or bilingual versions. Resolve major argument changes separately; translation follows the confirmed source and does not supply missing study facts. ## Prepare the source and constraints Provide complete passages, an existing glossary, the target language and intended use. Preserve section titles, citation markers, figure and table numbers, statistical units and registration details. For an abstract, include the journal's current format requirements. > Translate these results into English. Preserve numbers, units, P values, confidence intervals, citation markers and claim strength. Apply consistent terminology and list source uncertainties outside the manuscript instead of inventing corrections. ## Agree on terminology and compare protected content Record key terms, abbreviations, disease names and outcomes, then check consistency across the manuscript. Compare values and citations after translation; fluent language alone does not establish accuracy. | Content | Check | |---|---| | Numbers and statistics | Added, missing or altered values and units | | Citations and numbering | Correspondence with the source | | Claims | Whether uncertainty became certainty or association became causation | | Terminology | Consistent names for outcomes, variables and populations | Local protected-content checks can identify some additions or removals. They cannot reliably detect swapped group values, semantic drift or changes in claim strength; human comparison remains necessary. ## Deliver the translation and review trail Keep the source, translation, glossary, check report and unresolved questions. If further style work is needed, `deai` can improve repetitive prose while preserving facts, evidence and necessary disclosure. The [Chinese-to-English teaching example](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-translate/examples/worked-example-cn2en.md) presents the source, translation and separate issues. See [protected-content tool instructions](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-deai/references/protected-content-cli.md) for implemented checks. Continue with [manuscript sections](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/writing/) or [reviewer responses](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/reviewer-response/). ### Medical Manuscript Writing in IMRAD Format https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/writing/ Use this workflow when results are available and you need a first draft or a clearer discussion. Relevant skills include `writing-plan`, `introduction`, `methods`, `results`, `discussion` and `abstract`. Start with `manuscript-diagnose` if an existing draft needs assessment before editing. ## Prepare the inputs Provide the research question, actual study procedures, statistical outputs, figure versions, verified references and journal or thesis requirements. A results table can support planning; it cannot establish unknown methods or missing study facts. ## 1. Map the argument > Organize the confirmed question, main findings and supporting evidence. Plan the introduction, results and discussion paragraph by paragraph. Flag unsupported claims and do not draft the full paper yet. Check that primary and exploratory findings remain distinct. Keep conclusions within the study design's limits; observational association does not establish a causal mechanism. ## 2. Draft individual sections Methods follow actual execution records; results follow verified outputs. The discussion covers findings, comparison with prior evidence, interpretation, applicability and limitations. Prepare the title, abstract and keywords after the body is stable. > Draft the discussion first. Give each paragraph a clear role, preserve the result values and explain agreement or disagreement with available literature. List unverified citations as unresolved work. ## 3. Check the draft and hand it over Confirm that the body, tables, figures and abstract use the same result version. Check numerical and citation consistency, then review whether evidence supports the important claims. Deliver the draft with revision reasons and unresolved questions for supervisor or collaborator review. | Deliverable | Purpose | |---|---| | Argument map | Agree on the manuscript's main line | | Section drafts | Review and revise paragraph by paragraph | | Claim–evidence mapping | Trace the basis of important statements | | Missing-input list | Separate completed work from open questions | ## Teaching example The [statin cohort argument-map example](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-writing-plan/examples/示例-他汀类药物队列论证地图构建.md) demonstrates the progression from question to discussion plan. It is a synthetic example in Chinese, not a publication outcome. Continue with [medical translation](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/translation/) or [reviewer responses](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/tutorials/reviewer-response/). [Browse writing skills](https://duyunzhoudyz.github.io/dyz-med-paper-skills/en/skills/#writing) ### 项目状态、验证范围与使用许可 https://duyunzhoudyz.github.io/dyz-med-paper-skills/about/ ## 当前状态 DYZ MedPaper 是面向中国医生与医学研究生的公开测试版医学论文 Agent Skills 集合。核心内容包括 32 个技能、模板、参考资料、合成示例和本地脚本。 | 检查范围 | 能说明什么 | 不能说明什么 | |---|---|---| | 安装与文件自检 | 完整目录复制与字节一致性 | 宿主实际加载或调用正确 | | 离线回归与合成流程 | 明确实现的工具和文件合同行为 | 模型写作质量或医学结论正确 | | 文档站构建检查 | 页面、链接、语言和索引结构完整 | 搜索引擎已收录或排名提高 | 实际宿主路由、连续多轮写作和独立医学质量尚未全部完成验收。请查看 [实时离线检查](https://github.com/duyunzhoudyz/dyz-med-paper-skills/actions/workflows/check.yml)、[详细验证说明](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/docs/readme-reference.md#validation) 和 [发布记录](https://github.com/duyunzhoudyz/dyz-med-paper-skills/releases)。 ## 工具与访问条件 - 内置在线文献检索脚本覆盖 PubMed(题录核验同时解析 Crossref/PubMed 撤稿、关注声明与勘误提示)。知网、万方、维普及其他数据库需要合法访问权限;仅发现线索不能当作完成检索或全文获取。 - 正式统计执行需要批准的具体计划、R 或既定统计引擎及所需包;仓库另提供纯标准库的本地 PHI 初筛、基线 Table 1、KM 生存分析、二分类 Meta、诊断准确性与可溯源 SVG 矢量图脚本。 - 本地导出器支持基础 DOCX 审阅稿(`export_docx.py`)与 Word 原生修订留痕稿(`export_tracked_docx.py`,启用 ``);复杂排版、数学公式及最终视觉检查仍需 Word 或 LibreOffice。 - 本地数字、规范清单(6 套完整官方子项编号)或引用检查只覆盖明确实现的机械范围,不认证来源对论断的语义支持或规范合规。 ## 使用许可 项目源码公开,采用 [PolyForm Noncommercial License 1.0.0](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/LICENSE)。可在许可证允许的非商业范围内使用;超出许可范围需要[单独书面商业授权](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/COMMERCIAL-LICENSE.md)。第三方材料遵循各自条款。 ## 反馈与参与 维护者、公开源码和变更记录可在 [GitHub 项目主页](https://github.com/duyunzhoudyz/dyz-med-paper-skills) 查看。上方「当前状态」与「工具与访问条件」说明哪些检查由本地脚本完成、哪些仍依赖模型与人工判断。 在 [GitHub Issues](https://github.com/duyunzhoudyz/dyz-med-paper-skills/issues) 提供系统、宿主、复现步骤和脱敏报错。示例改进、术语修正和安装问题都可以反馈;请勿上传患者资料或账号凭据。 [贡献说明](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/CONTRIBUTING.md) · [维护与发布](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/docs/maintaining-discoverability.md) · [GEO 与机器可读层](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/docs/geo-playbook.md) · [项目介绍与引用文案](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/docs/discovery-kit.md) ### 医学论文写作与科研 Agent Skills https://duyunzhoudyz.github.io/dyz-med-paper-skills/ [安装并开始使用 →](https://duyunzhoudyz.github.io/dyz-med-paper-skills/install/){: .primary-link} [浏览 32 个技能](https://duyunzhoudyz.github.io/dyz-med-paper-skills/skills/){: .secondary-link} **32 medical writing and research skills for Claude Code and Codex.** 用中文讨论研究问题,按需要起草中文或英文稿件,把导师意见和审稿要求落实到具体段落。可以只处理眼前的一项任务,也可以围绕同一篇论文持续推进。 ## 从你现在最需要的一步开始 | 你手头有什么 | 可以先做什么 | 任务教程 | |---|---|---| | 一个临床问题 | 梳理检索范围,寻找已有研究和证据缺口 | [PubMed 与中文文献检索](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/pubmed/) | | 结果表和几篇文献 | 搭论证地图,再分节写引言、方法、结果和讨论 | [医学论文写作](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/writing/) | | 统计输出 | 区分统计显著性、效应大小与临床意义 | [统计结果解读](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/statistics/) | | 系统综述记录 | 管理筛选、纳排理由与 PRISMA 计数 | [系统综述工作流](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/systematic-review/) | | 一版中文稿件 | 翻译成英文并核对数字、引用和术语 | [医学论文中英翻译](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/translation/) | | 审稿意见与现有稿件 | 制作逐条回复表,定位每项正文修改 | [审稿意见回复](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/reviewer-response/) | ## 中文沟通,中英写作 这套技能围绕中国医生和医学研究生的科研日常设计:从科室病例与毕业课题出发,考虑现有数据、随访条件、导师意见、学校要求、目标期刊和版面费预算。研究思路可以先用中文讲清,再按用途整理成中英文稿件。 你不必记住技能名。比如: > 这是我的结果表、已核实文献和讨论初稿。请先梳理主要发现,再重写讨论,区分证据、合理解释和局限。缺依据的地方单列,并用中文说明修改理由。 已有初稿就从诊断和修订开始;收到审稿意见就进入修回。整篇论文不会在一次调用后自动完成。 ## 文件可以带走,过程可以继续 技能以 `SKILL.md`、模板、参考资料和本地 Python 脚本组织。论文材料存放在独立项目文件夹;写作提纲、章节、文献记录和修改清单可以继续交给导师与合作者审阅。内置导出器支持基础 Word 审阅稿。 [查看 GitHub 源码](https://github.com/duyunzhoudyz/dyz-med-paper-skills) · [项目状态与许可](https://duyunzhoudyz.github.io/dyz-med-paper-skills/about/) · [版本发布](https://github.com/duyunzhoudyz/dyz-med-paper-skills/releases) ## 先看清适用范围 当前为公开测试版。在线检索依赖数据库权限;内置在线脚本仅覆盖 PubMed。统计执行需要相应环境与具体分析计划的批准。真实宿主调用、连续写作和医学质量尚未全部完成独立验收。教学示例均为合成材料,不能作为发表或使用效果证明。 **DYZ MedPaper 是什么?** DYZ MedPaper 是一套面向中国医生与医学研究生的医学论文写作与科研 Agent Skills 集合,包含 32 个技能,覆盖选题、PubMed 文献检索、统计分析与结果解读、中英写作、选刊投稿与审稿修回。源码仓库为 github.com/duyunzhoudyz/dyz-med-paper-skills,文档站为 duyunzhoudyz.github.io/dyz-med-paper-skills。当前为公开测试版,采用 PolyForm Noncommercial 1.0.0 许可。 **可以在哪些智能体里使用?** 任何支持 Agent Skills 的宿主都可以加载完整技能目录:Claude Code 安装到 ~/.claude/skills,Codex 默认安装到 ~/.agents/skills,也可用 --target 指定其他宿主的技能目录。技能由 Markdown 说明、模板、参考资料和本地 Python 脚本组成,不绑定某个模型厂商;实际的任务路由与触发效果需要在目标宿主中核实。 **32 个技能分别覆盖什么?** 技能按五组组织:起点与项目状态 5 个、研究设计与文献证据 7 个、统计分析与图表 5 个、论证与论文写作 7 个、翻译修订与交付 8 个,合计 32 个。完整清单见 docs/skills.md;skills.json 还给出每个技能的中英标题、输入输出、触发词和指令地址。 **如何安装并开始第一项任务?** 支持三种零依赖安装方式:1)双击 installers/install-windows.bat 或 installers/install-macos.command;2)Claude Code 2.1.275+ 可用 /plugin install dyz-med-all --marketplace duyunzhoudyz/dyz-med-paper-skills 或按 5 个阶段轻量包安装(dyz-med-start/evidence/analysis/writing/delivery);3)终端运行 python scripts/install_codex.py [--pack all|start|evidence|analysis|writing|delivery] [--target 目录] 并加 --check 校验。需要 Python 3.10+。 **是否免费?许可如何?** 源码公开,采用 PolyForm Noncommercial License 1.0.0:许可证允许的非商业用途可以免费使用;超出许可范围需要单独书面商业授权,见仓库中的 COMMERCIAL-LICENSE.md。第三方材料遵循各自条款。 **能自动产出一篇可直接投稿的论文吗?** 不能。技能提供流程、模板和可追溯的检查点;正式统计分析、数据库访问、复杂排版和最终医学判断仍需要人和相应软件环境。内置在线检索脚本只覆盖 PubMed,知网、万方、维普等需要合法访问权限。教学示例均为合成材料,不是已发表研究或使用效果证明。 ### 在 Claude Code 与 Codex 安装医学论文 Skills https://duyunzhoudyz.github.io/dyz-med-paper-skills/install/ ## 1. 准备智能体与材料 准备支持 Agent Skills、已完成账号配置的智能体,以及 Python 3.10+。安装器只使用 Python 标准库,不安装模型、统计包或数据库账号。 另建一个论文项目文件夹,放置合成材料或获准处理的脱敏材料。本地脚本运行不代表整个 AI 会话离线;模型服务可能接收会话中的内容。 ## 2. 下载项目 在 [GitHub 仓库](https://github.com/duyunzhoudyz/dyz-med-paper-skills) 选择 **Code → Download ZIP**,解压并打开终端;也可使用 Git: ```bash git clone https://github.com/duyunzhoudyz/dyz-med-paper-skills.git cd dyz-med-paper-skills ``` Windows 可运行 `python --version`;使用 `py -3` 时替换下面的 `python`。macOS/Linux 通常使用 `python3`。 ## 3. 选择宿主并安装 ### Claude Code ```bash python scripts/install_codex.py --target "~/.claude/skills" python scripts/install_codex.py --target "~/.claude/skills" --check ``` ### Codex ```bash python scripts/install_codex.py python scripts/install_codex.py --check ``` 默认目标是 `~/.agents/skills/`。文件名沿用 `install_codex.py`,但可以用 `--target` 指定其他宿主的技能目录。成功时会显示 `PASS`、技能数量和安装路径。同名技能有差异时安装器会停止,避免覆盖本地修改。 ### 只给一个论文项目安装 将示例路径替换为已有项目目录。Claude Code 使用: ```bash python scripts/install_codex.py --target "/path/to/my-paper/.claude/skills" ``` Codex 使用: ```bash python scripts/install_codex.py --project "/path/to/my-paper" ``` 检查时保留相同参数并添加 `--check`。同一宿主的个人级与项目级安装二选一,避免重复。 ## 4. 开始第一项任务 在宿主中打开论文文件夹,重新加载技能后直接描述任务: > 请使用 dyz-med-orchestrator,看看当前论文材料,告诉我最该解决什么,先把这一步做好。需要补充的材料列出来。 Claude Code 可显式调用 `/dyz-med-orchestrator`;Codex 可显式调用 `$dyz-med-orchestrator`。安装自检验证文件一致性,实际加载情况仍须在宿主中核实。 ## 升级与排查 先运行带相同目标参数的 `--check`。需要升级时,从新版本源码运行安装器,并提供新的、位于源码与安装目录之外的备份目录: ```bash python scripts/install_codex.py --upgrade --backup "/path/to/new-backup" ``` Claude Code 需补上原来的 `--target "~/.claude/skills"`。不要只复制 `SKILL.md`;手动安装必须携带完整技能目录,以及 `LICENSE`、`NOTICE` 和 `COMMERCIAL-LICENSE.md`。 [完整安装维护说明](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/README.md#maintenance) · [浏览技能索引](https://duyunzhoudyz.github.io/dyz-med-paper-skills/skills/) · [选择任务教程](https://duyunzhoudyz.github.io/dyz-med-paper-skills/) ### 32 个医学论文 Agent Skills 完整索引 https://duyunzhoudyz.github.io/dyz-med-paper-skills/skills/ 按当前材料选任务,无需从头完成所有步骤。每个示例均为合成教学材料。多数技能正文与配套资料以中文为主;目录同时提供中英文任务词。 ### dyz-med-ideate — 医学科研选题 从临床观察梳理可验证的研究问题 - 输入 / Input: 临床观察、兴趣与资源限制 - 输出 / Output: 候选问题与可行性路线 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-ideate/SKILL.md ### dyz-med-init — 论文项目初始化 为已明确的研究建立论文项目与结构 - 输入 / Input: 研究问题、设计与已有材料 - 输出 / Output: 项目骨架与材料缺口 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-init/SKILL.md ### dyz-med-manuscript-diagnose — 医学稿件诊断 先评估已有稿件的问题与修改顺序 - 输入 / Input: 初稿、摘要或拒稿意见 - 输出 / Output: 诊断报告与修改优先级 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-manuscript-diagnose/SKILL.md ### dyz-med-orchestrator — 论文进度与下一步 根据现有材料判断下一步任务 - 输入 / Input: 论文文件夹与当前需求 - 输出 / Output: 进度地图与下一步建议 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-orchestrator/SKILL.md ### dyz-med-passport — 项目状态记录 内部维护已确认事实与产出版本 - 输入 / Input: 已确认事实、材料和批准记录 - 输出 / Output: 可追溯的项目状态 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-passport/SKILL.md ### dyz-med-mechanism — 机制与因果逻辑 检查机制假设、因果路径与证据层级 - 输入 / Input: 研究问题、设计与支持证据 - 输出 / Output: 机制链、假设缺口与验证路径 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-mechanism/SKILL.md ### dyz-med-data-adequacy — 数据充分性检查 判断数据能支持的问题和分析范围 - 输入 / Input: 数据字典、设计与获准数据 - 输出 / Output: 充分性报告与可分析范围 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-data-adequacy/SKILL.md ### dyz-med-registration — 研究注册核查 对照官方注册记录与方案、稿件 - 输入 / Input: 注册号、方案与稿件 - 输出 / Output: 注册对照表与待核实差异 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-registration/SKILL.md ### dyz-med-cnki — PubMed 与中文文献检索 设计 PubMed 和中文医学来源检索并记录覆盖范围 - 输入 / Input: 研究问题、数据库与纳排范围 - 输出 / Output: 检索式、记录与候选文献池 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-cnki/SKILL.md ### dyz-med-key-literature — 关键文献精读 对已有文献作结构化阅读与证据比较 - 输入 / Input: 候选文献与可访问全文 - 输出 / Output: 单篇简报与证据地图 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-key-literature/SKILL.md ### dyz-med-screening-pipeline — 系统综述文献筛选 管理去重、筛选、纳排理由与 PRISMA 计数 - 输入 / Input: 检索记录、筛选协议与全文 - 输出 / Output: 筛选记录、提取表与计数 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-screening-pipeline/SKILL.md ### dyz-med-evidence-audit — 论断与引用核查 检查文献和数据是否支持稿件论断 - 输入 / Input: 稿件、来源文献与研究结果 - 输出 / Output: 论断—证据矩阵与问题定位 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-evidence-audit/SKILL.md ### dyz-med-analyze — 医学统计分析 先形成计划,再按批准范围执行统计 - 输入 / Input: 获准数据、数据字典与研究问题 - 输出 / Output: 分析计划、脚本与结果记录 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-analyze/SKILL.md ### dyz-med-interpret — 统计结果与临床意义 区分统计证据、效应大小与临床重要性 - 输入 / Input: 统计结果与有来源的临床阈值 - 输出 / Output: 逐结局解释与谨慎措辞 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-interpret/SKILL.md ### dyz-med-sensitivity — 敏感性与稳健性分析 围绕明确的不确定性比较替代假设 - 输入 / Input: 主要结果、具体假设与批准范围 - 输出 / Output: 稳健性比较与局限说明 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-sensitivity/SKILL.md ### dyz-med-figures-tables — 医学论文图表 规划和核对 Table 1、森林图等图表 - 输入 / Input: 已执行分析输出与数据来源 - 输出 / Output: 图表、图注与一致性清单 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-figures-tables/SKILL.md ### dyz-med-flowchart — 研究流程图 由真实计数整理 PRISMA、CONSORT 等流程图 - 输入 / Input: 研究流程、筛选记录与计数 - 输出 / Output: 流程图、图注与计数核对 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-flowchart/SKILL.md ### dyz-med-writing-plan — 论文论证与写作提纲 把研究问题、结果与证据组织成段落计划 - 输入 / Input: 研究问题、已核实结果与文献 - 输出 / Output: 论证地图与章节提纲 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-writing-plan/SKILL.md ### dyz-med-introduction — 医学论文引言 围绕具体知识缺口组织引言 - 输入 / Input: 研究问题、写作提纲与核实文献 - 输出 / Output: 引言草稿与逻辑说明 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-introduction/SKILL.md ### dyz-med-methods — 医学论文方法部分 将实际研究过程写成可复核方法 - 输入 / Input: 方案、执行记录与分析计划 - 输出 / Output: 方法正文与缺项清单 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-methods/SKILL.md ### dyz-med-results — 医学论文结果部分 把核实后的统计输出写成结果正文 - 输入 / Input: 已核实统计输出、表格和图 - 输出 / Output: 结果正文与数字对应关系 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-results/SKILL.md ### dyz-med-discussion — 医学论文讨论 组织主要发现、证据比较、意义与局限 - 输入 / Input: 已核实结果、研究设计与文献 - 输出 / Output: 讨论草稿与论断边界 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-discussion/SKILL.md ### dyz-med-abstract — 标题、摘要与关键词 从已完成正文生成结构化摘要 - 输入 / Input: 完整正文、核实结果与期刊要求 - 输出 / Output: 标题、摘要与关键词 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-abstract/SKILL.md ### dyz-med-write — 全文初稿与统稿 按明确请求整合材料或统一已有章节 - 输入 / Input: 已确认研究材料或完成的章节 - 输出 / Output: 完整初稿或统稿版本 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-write/SKILL.md ### dyz-med-translate — 医学论文中英翻译 翻译正文与回复并保护数字、引用和术语 - 输入 / Input: 原文、术语表与目标语言 - 输出 / Output: 译文、术语表与保护检查 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-translate/SKILL.md ### dyz-med-deai — 学术语言自然化 改善重复空泛表达并保留事实和作者语气 - 输入 / Input: 原稿与表达偏好 - 输出 / Output: 自然化稿件与修改说明 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-deai/SKILL.md ### dyz-med-guideline — 医学研究报告规范 核对 STROBE、CONSORT、PRISMA 等适用条目 - 输入 / Input: 稿件、研究类型与适用规范 - 输出 / Output: 条目覆盖表与定位建议 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-guideline/SKILL.md ### dyz-med-review — 投稿前模拟审稿 从证据、方法等视角评估稿件 - 输入 / Input: 接近完成的稿件与附件 - 输出 / Output: 模拟审稿报告与修改建议 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-review/SKILL.md ### dyz-med-revision — 改稿与审稿意见回复 逐条落实修改并记录正文位置 - 输入 / Input: 稿件、审稿意见与已完成改动 - 输出 / Output: 修订稿、逐条回复与修改定位 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-revision/SKILL.md ### dyz-med-journal-finder — 医学论文选刊 按主题、读者、预算与当前要求筛选期刊 - 输入 / Input: 稿件主题、文章类型、预算与时间 - 输出 / Output: 分层候选期刊与核实依据 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-journal-finder/SKILL.md ### dyz-med-export — Word 审阅稿导出 整理章节、引用和图表生成基础 DOCX - 输入 / Input: 完成的章节、图表与文献 - 输出 / Output: 基础 DOCX 审阅稿与导出报告 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-export/SKILL.md ### dyz-med-submit — 投稿材料准备 准备投稿信、题名页与声明等本地文件 - 输入 / Input: 稿件、作者确认信息与期刊要求 - 输出 / Output: 投稿包与待确认清单 - 说明 / Instructions: https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-submit/SKILL.md 数据库检索、统计执行和复杂排版分别需要相应访问权限、环境和工具。`passport` 供工作流内部维护状态;查看进度时优先使用 `orchestrator`。 [安装完整技能目录](https://duyunzhoudyz.github.io/dyz-med-paper-skills/install/) · [能力与验证范围](https://duyunzhoudyz.github.io/dyz-med-paper-skills/about/) ### PubMed 与中文医学文献检索工作流 https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/pubmed/ `dyz-med-cnki` 是文献检索技能的目录名,职责包括 PubMed 与中文医学来源。内置在线脚本覆盖 PubMed;知网、万方、维普等需要相应工具和访问权限。 ## 准备检索问题与范围 给出研究人群、暴露或干预、比较、结局、研究类型,以及时间、语言和数据库范围。综述项目还需要预先规定的检索协议与纳排标准。 > 请围绕这个临床问题设计 PubMed 和中文数据库检索方案,列出核心概念、同义词和需要核实的主题词。记录每个来源实际访问的情况,无法执行的来源单列。 ## 执行与留痕 1. 先确认概念和检索式,再在有权限的数据库中执行。 2. 保存数据库名称、检索日期、实际查询式、筛选条件和返回记录数。 3. 导出数据库原生记录,保留来源;内置导入工具支持 RIS、字面值 BibTeX、NBIB 和 PubMed XML。 4. 去重时保留合并关系;候选文献发现与系统综述的正式筛选分别记录。 仅找到网页线索时,应标记为 `discovery_only`;不能表述成已完成数据库检索。只有摘要时,不能声称完成全文精读。 ## 从候选池到关键证据 让 `dyz-med-key-literature` 对可访问全文做结构化阅读:研究设计、人群、效应估计、局限和与当前问题的关系。除了支持研究,也保留矛盾研究和适用性差异。 > 请从已核实的候选文献中挑选关键研究。区分指南、系统综述、原始研究与矛盾证据,说明哪些已读全文、哪些只有摘要。 交付包括检索方案、执行记录、候选池、去重记录和文献简报。[文献检索教学示例](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-cnki/examples/worked-example.md) 为合成材料;[本地文献工具说明](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-cnki/references/literature-cli.md) 给出实际脚本用法。 开展系统综述时继续到 [筛选与 PRISMA 工作流](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/systematic-review/);已有可靠证据时可进入 [论文写作](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/writing/)。 ### 医学论文审稿意见逐条回复与修回 https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/reviewer-response/ 使用 `dyz-med-revision` 修订现有稿件、准备逐条回复或适配新期刊。收到拒稿信但不清楚根本问题时,可以先让 `manuscript-diagnose` 做只读诊断。 ## 准备材料 提供当前稿件、编辑信、按审稿人区分的完整意见、目标期刊要求和已完成的修改。保留稿件版本及页码或行号,便于回复定位。 > 请把审稿意见逐条编号,先形成修订计划。区分可以直接修改的文字、需要澄清的研究事实和需要重新分析的内容。起草回复时只描述实际已完成的改动。 ## 建立回复矩阵 | 字段 | 内容 | |---|---| | 意见编号与原文 | 保持与编辑信一一对应 | | 问题与处理方式 | 解释、补充文字、重新分析或有依据地说明不采纳 | | 正文修改 | 具体修改内容及稿件位置 | | 回复草稿 | 回应问题、解释依据和指向修改 | | 状态 | 已完成、待确认、待执行及原因 | 不要把“将补做分析”写成“已完成分析”。涉及新数据、重新分析、伦理、作者信息或其他源事实时,先获得所需确认,再记录实际执行结果。 ## 写出有定位的回复 回复应直接回应问题,说明具体改动和位置。不同意某项要求时,提供与研究设计和证据相关的理由,并考虑正文是否需要补充解释。避免整封信只有礼貌套话。 > 根据已完成的正文修改,更新逐条回复。每条注明修改位置;没完成的地方保留待办。核对回复承诺与正文是否一致,再给出本轮修回仍缺什么。 ## 最后核对与交付 交付修订稿、回复信、修改记录和待办清单。再检查摘要、结果、讨论和图表的一致性。`submit` 可以整理本地投稿材料;外部上传或正式提交需明确授权与最终确认。 [逐条回复教学示例](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-revision/examples/worked-example-reviewer-response.md) 为合成案例,展示修改定位与未完成工作记录。[查看修订与交付技能](https://duyunzhoudyz.github.io/dyz-med-paper-skills/skills/#delivery) · [需要英文回复时](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/translation/) ### 医学统计结果解读与临床意义 https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/statistics/ 医学统计相关任务由不同技能处理:`data-adequacy` 判断数据能支持什么,`analyze` 制定并执行获批计划,`interpret` 解释已有结果,`sensitivity` 检查具体假设变化下的稳健性。 ## 先选对起点 | 现状 | 适合交代的任务 | |---|---| | 只有原始数据 | 先检查结构、缺失和数据充分性 | | 需要主分析 | 形成可审阅计划,确认后执行 | | 已有统计输出 | 解释效应、区间和临床重要性 | | 担心某项假设 | 明确不确定性,再安排敏感性分析 | ## 解释已有结果 准备结果表、效应尺度、单位、参照组、置信区间、分析人群和模型说明。讨论临床重要性时,还需提供预先规定且有来源的阈值;缺少阈值就记录缺口。 > 解释这份现有结果,不重新建模。逐个结局说明效应大小和区间,区分统计证据与临床意义。临床阈值没有依据时请明确写出,不自行补造。 核对解释有没有遗漏方向、单位或参照组。避免把“不显著”直接解释为“没有效应”,也不要仅根据 P 值替代对效应估计和不确定性的讨论。 ## 需要补充分析时 先指出具体疑问,例如缺失数据处理或分析人群变化。把替代假设、需要的数据、预期输出和执行条件写进计划,再按批准范围运行。保留脚本、日志、诊断与结果版本。 ## 交付与写作衔接 输出逐结局解释表、需要确认的临床阈值、谨慎表述及待补分析。`results` 根据核实数字写结果;`discussion` 根据研究设计和证据讨论意义;`figures-tables` 核对图文一致性。 [结果解读教学示例](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-interpret/examples/worked-example.md) · [统计分析示例](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-analyze/examples/worked-example-cohort.md) · [论文写作教程](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/writing/) 示例为合成材料。技能输出需由研究者和统计专业人员审阅,不能据此认证研究结论。 ### 系统综述文献筛选与 PRISMA 流程图 https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/systematic-review/ 适合已有检索协议和数据库记录的系统综述、Meta 分析或范围综述。`screening-pipeline` 管理筛选证据链,`flowchart` 整理流程图,`guideline` 核对适用报告规范。 ## 需要哪些输入 准备检索协议、纳排标准、各数据库原生导出、已获取全文和筛选安排。没有覆盖到的来源应明确标记;仅有 PubMed 检索不能描述为已完成多数据库检索。 > 导入这些检索记录,保留来源和去重依据。依据这份协议建立标题摘要与全文筛选表,保留独立判断、分歧处理和全文排除理由。先列出缺失材料。 ## 保留三个层次的记录 | 层次 | 要记录什么 | |---|---| | 检索记录 | 每个来源的查询、日期、返回数与重复关系 | | 报告 | 题录、全文获取状态、筛选决定与排除理由 | | 研究 | 同一研究的多个报告关联、数据提取与偏倚评估 | 筛选过程中保留独立判断与分歧处理记录。不同报告可能来自同一研究,不能简单把“论文篇数”当作“研究数”。缺失全文和待判定项目需要独立记录。 ## 生成并核对流程图 先核对各阶段分支与总数,再交给 `flowchart` 生成 PRISMA 图稿和图注。工具可帮助检查计数关系,但不能替代筛选判断。 > 根据已完成的筛选记录整理 PRISMA 计数,分别报告 records、reports 和 studies。标明尚未完成的筛选或全文获取步骤,不把待办写成完成。 ## 何时进入 Meta 分析 在数据提取、研究关联与适用偏倚评估明确后,才讨论合并模型与敏感性分析。正式统计由 `analyze` 按获批计划和实际工具环境执行;技能安装不包含统计软件和包。 [筛选教学示例](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-screening-pipeline/examples/worked-example.md) · [流程图示例](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-flowchart/examples/worked-example.md) · [返回检索教程](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/pubmed/) 上述示例为合成材料,不能作为已经完成的真实系统综述记录。 ### 医学论文中英翻译与数字、引用保护 https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/translation/ 使用 `dyz-med-translate` 处理中文到英文、英文到中文或双语版本。需要重构论证时先修改内容;翻译阶段依据已确认原稿,不补写原文缺失的研究事实。 ## 准备原文与约束 提供完整待译段落、已有术语表、目标语言和用途。尽可能保留章节标题、引用标记、图表编号、统计单位及注册信息。摘要另提供期刊当前格式要求。 > 将这些结果段落翻译为英文,保留数字、单位、P 值、置信区间、引用标记和结论强度。统一术语。原文疑点请在稿外列出,不在译文里替我修造事实。 ## 先确认术语,再核对受保护内容 记录关键术语、缩写、疾病名称和结局名称,检查同一概念在全文中是否一致。翻译后逐项对照数字与引用,不把语言自然化等同于内容准确性。 | 内容 | 核对重点 | |---|---| | 数字与统计量 | 是否新增、丢失或更换数值和单位 | | 引用与编号 | 标记与原稿是否对应 | | 结论表述 | 是否把可能性改为确定性或关联改为因果 | | 术语 | 同一结局、变量和人群是否采用一致名称 | 本地保护检查可发现部分增删差异,但不能可靠识别两组数字互换、语义漂移或结论强度变化,仍需人工对照。 ## 交付哪些文件 保留原文、译文、术语表、检查报告和稿外疑问。需要进一步润色时,再使用 `deai` 改善模板化或重复语言,继续保留事实、证据和必要披露。 [中译英教学示例](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-translate/examples/worked-example-cn2en.md) 展示原文、译文和需要另外处理的问题;[受保护内容工具说明](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-deai/references/protected-content-cli.md) 说明本地检查范围。 完成译文后可返回 [章节写作](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/writing/) 或继续 [审稿回复](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/reviewer-response/)。 ### 医学论文写作:从结果表到 IMRAD 章节 https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/writing/ 适合已有研究结果、准备写论文或重整讨论的医生与医学研究生。相关技能:`writing-plan`、`introduction`、`methods`、`results`、`discussion`、`abstract`。已有稿件但不清楚问题时,先用 `manuscript-diagnose`。 ## 准备哪些材料 准备研究问题、实际执行方案、统计结果表、图表版本、核实后的文献,以及目标期刊或毕业论文要求。只提供结果表时,可以先做提纲;不能据此补写未知研究过程。 ## 第一步:建立论证地图 > 根据这些已核实材料,先整理研究问题、主要发现和每个发现的证据。给我引言、结果和讨论的段落计划,标出哪些论断缺依据。暂时不要生成整篇论文。 检查提纲是否区分主要结局与探索性结果,是否把结论限制在研究设计能支持的范围内。观察性关联不应直接写成因果机制。 ## 第二步:分节起草 方法部分依据实际执行记录;结果部分使用已核实输出;讨论部分组织主要发现、既有证据比较、合理解释、适用性与局限。正文稳定后再整理标题、摘要和关键词。 > 请先写讨论。每段只承担一个主要任务,保留结果数值,对照已有文献解释异同。没有核实的引用列为待办,不填造参考文献。 ## 第三步:核对并交付 核对正文、图表和摘要是否采用同一结果版本;再检查数字、引用及重要论断的证据。把草稿、修改理由和待补材料一起交给导师或合作者。 | 交付 | 用于什么 | |---|---| | 论证地图 | 确认文章主线和章节分工 | | 分节草稿 | 逐段讨论与修改 | | 论断—证据对应 | 回溯重要说法的依据 | | 待补问题清单 | 区分完成内容与未解决事项 | ## 仓库中的教学示例 [他汀类药物队列论证地图](https://github.com/duyunzhoudyz/dyz-med-paper-skills/blob/main/dyz-med-writing-plan/examples/示例-他汀类药物队列论证地图构建.md) 展示如何从研究问题、主要发现走到讨论段落。它是合成教学示例,不是已发表成果。 下一步可进行 [中英翻译](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/translation/) 或 [审稿意见回复](https://duyunzhoudyz.github.io/dyz-med-paper-skills/tutorials/reviewer-response/)。[查看全部写作技能](https://duyunzhoudyz.github.io/dyz-med-paper-skills/skills/#writing)