Biomedical Paper Reader skill
Reads one biomedical paper figure by figure: why each experiment was run, what the data support, what to borrow
- 156 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill reads papers and writes a report, running nothing on the system
- Uploaded PDFs and text go to the provider of the agent model
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/biomedical-paper-readerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill reads papers and writes a report, running nothing on the system
- Uploaded PDFs and text go to the provider of the agent model
This is third-party code. Review the repository files before installing.
What it does
Biomedical Paper Reader is a skill for deep reading of a single biomedical paper. The input is a PDF, DOI, link or text, and the output is a reader report: research question, methods and purpose, results, how the argument connects, contribution, scientific limits and ideas to borrow for your own study design. The report defaults to Chinese, with English and bilingual versions on request. There are three modes. A full read covers the available text and methods, scans main and supplementary figures and goes deeper into panels when needed. A quick read starts from the abstract and result structure, then checks the figures that carry the conclusion and lists what was not read. The third mode takes one figure or method with its context and controls. Working records (source version, viewing events, coverage table, evidence index) are kept apart from the report and shown on request. The skill describes an image only after actually viewing it, never writes an observation as causation and distinguishes analysis units such as a cell versus a donor. A local desktop app, BioReader for macOS and Windows, ships alongside.
Who it is for. Biologists, clinician researchers and postgraduates who read papers outside their narrow field and want to understand the evidence, not only a summary.
Good fit when
- Before a journal club you need a breakdown of one paper explaining why each experiment was run
- You need to decide quickly whether a paper is worth reading in full and get a list of unchecked places
- You want to see how the author moved from a figure to a conclusion and where the evidence ends
- You are looking for a good experimental design to adapt for your own work
Not a fit when
- You need a full translation of a paper: that is not the main task of the skill
- You need a systematic review or a comparison of several papers: the skill is built for one paper
- You need a clinical recommendation: the report explains the evidence of a paper and does not replace a physician decision
Example request
Do a full read of this paper from the PDF: for each figure explain why the experiment was run and what it showsLimitations
The report relies on the available text and figures. If you only have a fragment or a secondary breakdown, the skill works with it but states the coverage limits and marks judgments as provisional. Supplementary material is taken only from local files, and downloads from the network happen on explicit request.
Reading figures needs a model with image support. The author describes an abort problem in long answers from reasoning models and its fix in version 0.13.1, but watch for truncation on long papers. The user controls the report language: an English paper does not switch the output to English by itself. Desktop releases can lag behind the `main` branch.
How to disable. Delete the `biomedical-paper-reader` folder from the agent skills directory, for example `~/.codex/skills`. The BioReader desktop app is removed like any program and reports stay in your folders.
FAQ
How do I install the skill?
Clone the repository into the agent skills directory, for example `git clone https://github.com/Gaoyuan-0423/biomedical-paper-reader.git ~/.codex/skills/biomedical-paper-reader`.
What input does the skill accept?
A PDF, DOI, paper link or full text. It also works with a secondary breakdown, marking the coverage as limited.
How does quick read differ from a full read?
A quick read starts from the abstract and result structure, checks only the key figures and lists what was not read. A full read goes through all main and supplementary figures.
Is there an app with an interface?
Yes, BioReader for macOS and Windows. Download it from the repository releases page; it runs locally next to the skill.
Can I get an English report?
Yes, on request. The user picks the language, and a bilingual report shares one evidence set, so numbers and strength of conclusions match.
Discussion
No questions or reviews yet.
Related
SkillsA skill for deep-reading one scientific paper: six sections, key figures in the text and each claim checked against the source.
Two skills for Claude Code and Codex: the agent translates papers in the EasyRead library and answers your margin notes
paper-reading-skill turns one arXiv AI or robotics paper into a report with each claim tied to evidence in the source.
A skill that searches five open databases for papers and builds an offline report with relevance, BibTeX and search strategies.
A skill for reading research papers in Chinese: it first states what was read, then ties every number to a table or figure.
Skills for clinical research: literature search, statistics, reporting guideline and citation checks, journal submission
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Biomedical Paper Reader (this card) | no | 156 | |
| kelip-paper-reading | no | 355 | reads papers in general without biomedical focus or a figure mode |
| Medical Research Agent Skills | no | 2,047 | a set of medical research skills, with no figure-based review of one paper |
README in short
The skill builds a reader report around the questions "why was the experiment designed this way", "what do the data show" and "what can be borrowed". The author separates the report from working records, requires describing an image only after viewing it and matches wording to the strength of a conclusion. The repository contains sample reports and acceptance results.
README badge
Original SKILL.md
--- name: biomedical-paper-reader description: "为单篇生物医学研究论文生成中文精读报告：逐图解释实验为什么这样设计、方法与结果如何支持结论、证据边界在哪里，并提炼可借鉴的研究设计；支持速读筛选与指定图／方法专题解析，英文或中英双语按需提供。输入可为 PDF、DOI、链接或正文。Deep-read, figure-by-figure analysis, quick screening, or focused figure/method analysis of one biomedical paper (PDF, DOI, link, or full text). 全文翻译、系统综述与多篇比较不是默认任务。" --- # 生物医学文献阅读 围绕单篇论文回答：作者为什么这样设计，每一步证据支持到哪里，哪些设计可以借鉴。版本 v0.13.1（[更新记录](CHANGELOG.md)）。 默认交付是供读者理解论文的**阅读报告**： - 正文由研究问题、方法与目的、结果、论证衔接、贡献、科学局限和研究借鉴构成。 - 来源版本、查看事件、覆盖表、证据索引等工作记录独立保存，不进入正文，也不默认生成可读核查附件；分离规则见[证据规则](references/evidence-rules.md)。 - 用户明确要求核查、审稿或复现审计时，再按其目的展示相应记录。 - 正文保留自然的图号／面板引用，以及真正影响具体判断的来源或材料限制。 ## 选择语言与模式 - 默认中文，未指定时直接开始，不反复询问；论文语言不改变默认。 - 明确英文时从证据直接写成自然学术英文，不先写中文再逐句翻译；明确双语时两种表述共用证据记录，数字、方向、否定和判断强度必须一致。细则见[模式与语言适配](references/reading-modes.md)。 - 默认完整精读；“速读／筛选／quick read”选速读；“只解析 Figure 3／讲解这个方法”选专题解析。 - 语言只改变表达，模式改变实际读取范围；两者都不降低范围内的证据标准。 - 用户表达足够明确就开始，不强制填写参数。 ## 参考文件加载 | 时机 | 读取 | |---|---| | 开始前 | [模式与语言适配](references/reading-modes.md)、[证据规则](references/evidence-rules.md) | | 识别研究设计后 | [研究检查点](references/study-checks.md) 中相关段落 | | 写报告时 | [报告模板](references/report-template.md) | | 处理文件、PDF 或图像时 | [材料处理与执行规则](references/material-handling.md) | | 有图任务 | [逐图阅读记录](references/reading-record.md) | ## 1. 确认身份与范围 - 输入可以是 PDF、DOI、链接、原文片段或二手解读；优先用户原文，只有标识或二手解析时才尝试获取可访问正文。 - 核对标题、DOI、来源版本；能访问官方记录时检查更正、更新或撤稿提示。文件日期不能证明更正已纳入，细则见证据规则的来源覆盖。 - 补充材料默认只用本次上传、论文同目录或用户指定目录中可确认属于该论文的本地文件；不遍历无关目录，没有附件时继续可完成的正文分析，只有用户明确要求补取才联网下载。执行细则见[材料处理](references/material-handling.md)。 - 完整精读对已有主图、补图及 Extended Data 逐张概览并联读完整图注；速读与专题可选图组，但必须包含所依赖的对照与验证。 - 补充表和数据文件按相关问题读取，不要求复算全部数据。 - 记录正文、Methods、图像、图注、补充材料、数据和代码的实际读取范围。 - PDF 区分文件页序与印刷页码；HTML 用真实小标题和图号。 - 身份不明且阻碍继续时才澄清；没有研究方向时给条件式适用场景。 - 仅有片段则交付有范围声明的暂定分析；缺材料不等于作者未做实验或论文质量低。