duyunzhoudyz/dyz-med-paper-skills

What it is

DYZ MedPaper packages 32 Agent Skills aimed at Chinese doctors and medical graduate students who need to move a clinical or thesis project from research question to submittable manuscript. The skills cover topic framing, PubMed literature searching, Chinese and English drafting, statistics interpretation, journal selection, and responding to reviewer comments. Skills follow the Agent Skills SKILL.md directory layout and can be installed into hosts such as Claude Code and Codex, which provide skill discovery, tool permissions, and invocation. See the 32-skill index and the Chinese documentation site.

Quick start

1. Prepare an agent that supports Agent Skills and has the required account configuration, plus Python 3.10+ for the installer and local checks.

2. Get the code and enter the directory:

   git clone https://github.com/duyunzhoudyz/dyz-med-paper-skills.git
   cd dyz-med-paper-skills
   

3. Install for Claude Code (personal skill directory):

   python scripts/install_codex.py --target "~/.claude/skills"
   python scripts/install_codex.py --target "~/.claude/skills" --check
   

   Or for Codex, which defaults to ~/.agents/skills/:

   python scripts/install_codex.py
   python scripts/install_codex.py --check
   

   On Windows with py -3, replace python accordingly; on macOS/Linux use python3. The installer copies all 32 skills and license files and prints PASS with the skill count and install path. It does not access the network or install models or R/Python packages. See the installation section.

4. Open a paper project folder in the agent, place your materials inside, and ask in plain language. A starting prompt is:

   请使用 dyz-med-orchestrator，看看这个文件夹里的论文材料，告诉我现在最该解决什么，先帮我把这一步做好。需要补充的材料列出来。

   Explicit invocation: /dyz-med-orchestrator in Claude Code or $dyz-med-orchestrator in Codex. Review the agent's task understanding and gaps before continuing. The --check run only verifies installed files; whether skills actually load must be confirmed inside the host.

Constraints

• Literature retrieval needs network access, tools, and database permissions. The built-in online search tool covers PubMed only; multi-database systematic reviews need other sources.
• Formal statistical work needs appropriate software and an analysis plan you confirm. Reading Excel, PDF, or Word files requires the corresponding tools. Built-in Word export suits basic review drafts; complex formatting, equations, and tracked changes need separate tools.
• Local scripts do not make the whole AI session offline. Model services may receive session content, so do not submit identifiable patient information, hospital credentials, or unapproved materials.
• The project is a public beta. Continuous writing performance, behavior across hosts, and independent medical quality evaluation are not fully validated.
• The source is public under PolyForm Noncommercial License 1.0.0; uses beyond the license require separate written commercial authorization.

Repository notes

The repo includes synthetic teaching examples rather than user feedback or published papers, including building an argument map from results, diagnosing a rejected manuscript, Chinese-to-English result paragraph translation, and reviewer-response drafting. The executable workflow document describes how literature organization, text checking, translation, and Word export connect.

HotRepo's take

HotRepo recorded 164 stars on 2026-11-04 in the Python new-project list, with 49 stars gained over a window beginning 2026-11-03 14:45 UTC. The hourly star samples rose from 100 on 2026-11-03 01:45 UTC to 164 by 2026-11-04 14:45 UTC. There is no daily all-language rank or Hacker News thread to draw on. The repository's own materials are explicit that this is a public beta and that independent medical quality evaluation is incomplete, so the numbers indicate recent attention rather than proven writing outcomes.