简体中文 | English
A modular AI skill suite for academic researchers, built for Claude Code. Dr. Cook covers the full research lifecycle — from literature review and gap analysis through manuscript writing, peer review, bioinformatics analysis, and grant applications — with particular depth in TCM (Traditional Chinese Medicine) and Chinese-language academic contexts.
Author: Harold W. Chen · biochen.org · chenwen@biochen.org
Dr. Cook provides 16 specialized modules that work independently or chain together as pipelines. A central router classifies your intent and loads the right module automatically. Context (text, parameters, findings) passes between pipeline stages without re-pasting.
| Category | Module | What It Does |
|---|---|---|
| Writing | paper-writer |
Draft full manuscripts or individual IMRaD sections |
grant-writer |
Write NSFC grant applications (面上/青年/重点项目) | |
rebuttal-writer |
Draft point-by-point reviewer response letters | |
cover-letter-writer |
Write journal submission and revision cover letters | |
| Literature | literature-reviewer |
Synthesize literature and produce structured reviews |
research-gap-finder |
Identify methodological, population, and knowledge gaps | |
citation-checker |
Validate citation format and detect in-text mismatches | |
literature-downloader |
Retrieve full-text articles via PubMed MCP | |
| Language | english-polisher |
Three-pass English polish (grammar → register → fluency) |
chinese-polisher |
Three-pass Chinese polish for academic publication | |
en-to-zh |
English-to-Chinese academic translation | |
zh-to-en |
Chinese-to-English academic translation | |
| Review | paper-reviewer |
Self-check or peer review mode with referee report |
| Analysis | bioinformatics-assistant |
DEG, enrichment (GO/KEGG/GSEA), and network pharmacology |
data-visualizer |
Generate ggplot2/matplotlib code for publication figures | |
method-designer |
Design CRISPR, multi-omics, and network pharmacology experiments |
| Command | Sequence |
|---|---|
/cook grant |
literature-reviewer → research-gap-finder → grant-writer |
/cook submit |
paper-writer → citation-checker → english-polisher → cover-letter-writer |
/cook review self |
paper-reviewer → english-polisher → citation-checker |
/cook review peer |
paper-reviewer (referee report mode) |
/cook pipeline: A → B → C |
Any custom module sequence |
| Domain | Citation Format | Notes |
|---|---|---|
tcm |
Vancouver | Bilingual EN/ZH terms; syndrome differentiation; NSFC grant support |
bioinformatics |
Vancouver / APA | HGNC gene names; code availability; GEO/SRA data deposition |
clinical |
Vancouver | CONSORT / STROBE / PRISMA; ethics statement; trial registration |
pharmacology |
APA | IUPHAR nomenclature; dose units; in vitro vs in vivo distinction |
- Claude Code (any recent version)
- For
literature-downloader: the PubMed MCP server configured in Claude Code
Clone this repository directly into your Claude Code skills directory:
git clone https://github.com/haroldwchen/dr-cook ~/.claude/skills/dr-cookThat's it. Claude Code discovers skills automatically from ~/.claude/skills/.
Manual install: Download or copy this repository and place the folder at ~/.claude/skills/dr-cook/. The folder name must be dr-cook.
Invoke Dr. Cook from any Claude Code session:
dr-cook
The router displays a help menu with available commands and pipelines. Type your request in plain English or Chinese — Dr. Cook classifies your intent and loads the appropriate module.
Each module can also be invoked directly by name, bypassing the router:
dr-cook:paper-writer
dr-cook:english-polisher
dr-cook:bioinformatics-assistant
Write the Methods section for my TCM RCT targeting Chinese Medicine journal.
帮我写一个关于针灸治疗高血压的NSFC面上项目申请书。
Run a DEG analysis on my RNA-seq count matrix using DESeq2.
Polish this abstract for submission to Nature Medicine.
/cook submit
After any module completes, follow-up requests automatically carry context forward without re-pasting:
[english-polisher completes on your abstract]
You: Now translate it to Chinese. ← automatically picks up the polished text
dr-cook/
├── SKILL.md # Main router
├── shared/
│ ├── context-schema.md # JSON data contract between modules
│ └── domain-configs.md # Per-domain defaults (citation, terminology)
├── modules/
│ ├── writing/ # paper-writer, grant-writer, rebuttal-writer, cover-letter-writer
│ ├── literature/ # literature-reviewer, research-gap-finder, citation-checker, literature-downloader
│ ├── language/ # english-polisher, chinese-polisher, en-to-zh, zh-to-en
│ ├── review/ # paper-reviewer
│ └── analysis/ # bioinformatics-assistant, data-visualizer, method-designer
├── tests/
│ ├── router-test-cases.md # 40-case router intent classification tests
│ ├── pipeline-integration-tests.md
│ └── module-tests/ # 3–5 test cases per module
└── docs/
└── requirements.md # Full requirements specification
Each module follows this layout:
modules/<category>/<module-name>/
├── SKILL.md # Module instructions (800–1200 words)
└── references/ # Domain knowledge files (≤800 words each)
Dr. Cook is built as a Claude Code skill — a Markdown instruction set that Claude reads at runtime. There is no compiled code. When you invoke a module, Claude loads the relevant SKILL.md and references/ files via its Read tool and executes the workflow described there.
Context passing: All modules share a standardized context_output JSON schema. When modules run in a pipeline, each stage reads the previous module's output (text, parameters, key findings) without requiring the user to re-provide information.
Bilingual support: Every module handles English and Chinese input. The router intent table contains EN and ZH trigger keywords for all 16 modules.
Contributions welcome. Please keep pull requests focused:
- New modules: Follow the SKILL.md template (800–1200 words; required sections: Overview, Parameters, Workflow, Output Format, context_output, References). Add 3–5 test cases in
tests/module-tests/. - Reference file updates: Keep each file ≤800 words. No content duplication across files.
- Bug reports: Describe the input, the module invoked, and what the output was vs. what was expected.
MIT License — see LICENSE for details.
Copyright (c) 2026 Harold W. Chen · biochen.org