A Claude Code skill that retrieves scientific papers from PubMed and creates accessible, plain-language research summaries.
What It Does
This skill bridges the gap between complex scientific literature and general understanding by:
- Retrieving - Searches PubMed for relevant scientific papers on any biomedical topic
- Analyzing - Identifies key themes, findings, and implications from multiple papers
- Translating - Converts technical research into plain language that anyone can understand
- Summarizing - Creates comprehensive markdown reports with clear structure and context
When to Use
The skill automatically activates when you:
- Ask about medical or scientific research (e.g., "What's the latest on diabetes treatment?")
- Request information about diseases, treatments, or clinical trials
- Want to understand scientific consensus on health topics
- Need to catch up on recent research in a specific area
Quick Start
Simply ask natural questions:
Summarize research on immunotherapy for lung cancer
What does research say about ketogenic diet for epilepsy?
Tell me about recent CRISPR gene therapy studies
Claude will automatically:
- Use the skill to search PubMed
- Analyze the most recent papers
- Create a comprehensive plain-language summary
- Save it as a markdown file in your working directory
Output Format
Each summary includes:
- Overview: What the research area is and why it matters
- Key Research Themes: Major findings organized by topic
- Clinical Implications: What this means for patients and treatment
- Challenges: Current problems and limitations
- Key Terminology: Definitions of important scientific terms
- Selected Papers: Highlighted papers with PMIDs and links
- Summary: Overall synthesis and future directions
File Naming
Summaries are saved as:
Research_Summary_{Topic}_{YYYY-MM-DD}.md
Examples:
Research_Summary_Immunotherapy_Melanoma_2025-12-13.md
Research_Summary_Alzheimers_Treatment_2025-12-13.md
Requirements
The skill includes a built-in pubmed_search.py script located at .claude/skills/medical-research/pubmed_search.py. This script requires:
- Python 3.x
- Biopython library (
pip install biopython)
These should already be installed in the Vibe environment.
Customization
Retrieve More Papers
Edit .claude/skills/medical-research/pubmed_search.py and change the max_results parameter:
papers = search_pubmed(query, max_results=20) # Default is 10
Set Your Email
PubMed requires an email address for API access. Update in .claude/skills/medical-research/pubmed_search.py:
Entrez.email = "your.email@example.com"
Search Strategy
For better results:
Be Specific:
- ✓ "immunotherapy for triple-negative breast cancer"
- ✗ "cancer treatment"
Use Medical Terms:
- ✓ "myocardial infarction rehabilitation"
- ✗ "heart attack recovery"
Include Context:
- ✓ "CRISPR gene editing sickle cell disease"
- ✗ "CRISPR"
Examples
Example 1: Treatment Research
Input:
Summarize research on immunotherapy for breast cancer
What Happens:
- Searches PubMed for "immunotherapy breast cancer"
- Finds 14,190 papers, retrieves 10 most recent
- Identifies themes: PD-L1 paradox, resistance mechanisms, novel approaches
- Creates 3,000+ word plain-language summary
- Saves as
Research_Summary_Immunotherapy_Breast_Cancer_2025-12-13.md
Example 2: Disease Mechanism
Input:
What does research say about Alzheimer's disease causes?
What Happens:
- Skill activates automatically
- Searches for "Alzheimer's disease pathogenesis"
- Explains amyloid hypothesis, tau protein, neuroinflammation in accessible terms
- Discusses competing theories and current debates
- Saves comprehensive summary
Example 3: Emerging Technology
Input:
Tell me about mRNA vaccine technology
What Happens:
- Searches "mRNA vaccine mechanism clinical trials"
- Explains how mRNA vaccines work (using simple analogies)
- Covers COVID-19 applications and future potential
- Discusses safety, efficacy, and ongoing research
- Saves detailed summary
Tips for Best Results
Getting Comprehensive Summaries
- Be specific about what aspect interests you
- Mention if you want focus on mechanisms, treatments, or clinical trials
- Specify populations if relevant (pediatric, elderly, specific conditions)
Understanding Technical Topics
- The skill will define key terms in a "Key Terminology" section
- Explanations use analogies and plain language
- Complex mechanisms are broken down step-by-step
Finding Specific Papers
- Use the "Selected Papers" section for highlighted research
- Each paper includes PMID (PubMed ID) for direct lookup
- Links provided for one-click access to PubMed
Follow-Up Questions
After receiving a summary, you can ask:
- "Can you explain the [specific mechanism] in more detail?"
- "Which of these approaches is most promising?"
- "Are there clinical trials I can look into?"
Limitations
What This Skill Does:
- Summarizes published peer-reviewed research
- Explains scientific concepts in plain language
- Identifies trends and patterns in literature
- Provides educational information
What This Skill Doesn't Do:
- Provide medical advice or treatment recommendations
- Replace consultation with healthcare providers
- Guarantee accuracy of cited research
- Cover unpublished or non-PubMed indexed research
Currency Note:
- Focuses on most recent papers (typically last 1-2 years)
- Very new research (last few weeks) may not yet be indexed
- Some emerging topics may have limited papers available
Troubleshooting
"No papers found"
Try:
- Broadening search terms (use general terms)
- Checking spelling of medical terms
- Using alternative terminology (e.g., "myocardial infarction" vs "heart attack")
- Verifying the topic has biomedical research (PubMed covers life sciences)
"Papers are too technical"
The skill should automatically translate complex findings. If unclear:
- Ask for clarification on specific terms
- Request the skill focus on clinical implications
- Check the "Key Terminology" section for definitions
"Results seem narrow"
- Modify the
.claude/skills/medical-research/pubmed_search.py script to retrieve more papers
- Use broader search terms
- Ask for a follow-up search on related topics
Technical Details
How It Works
- Query Processing: Converts your natural language question into a PubMed search query
- Paper Retrieval: Uses Biopython's Entrez module to query PubMed API
- Analysis: Reads titles, abstracts, authors, and metadata from retrieved papers
- Synthesis: Identifies themes, patterns, and key findings across papers
- Translation: Converts technical language to plain English with analogies and context
- Structuring: Organizes information into clear sections with hierarchy
- Documentation: Saves as markdown file with proper citations
Data Sources
- Primary: PubMed/MEDLINE (https://pubmed.ncbi.nlm.nih.gov/)
- Coverage: 35+ million biomedical literature citations
- Update Frequency: Daily; new papers added continuously
- Access: Free, public database maintained by US National Library of Medicine
Quality Assurance
Summaries are designed to:
- Write at 10th-grade reading level (accessible to general audience)
- Define all technical terms and abbreviations
- Cite specific papers with PMIDs for verification
- Note limitations and areas of uncertainty
- Include educational disclaimers
- Present balanced view including challenges
Contributing
To improve this skill:
- Enhance the script: Modify
.claude/skills/medical-research/pubmed_search.py to retrieve more metadata
- Improve templates: Update
SKILL.md with better structure suggestions
- Add examples: Contribute sample summaries to
EXAMPLES.md
- Report issues: Note any topics that produce poor summaries
Related Resources
- PubMed: https://pubmed.ncbi.nlm.nih.gov/
- Biopython Tutorial: https://biopython.org/wiki/Documentation
- MeSH Browser: https://meshb.nlm.nih.gov/ (for finding medical search terms)
- Plain Language Guidelines: https://www.plainlanguage.gov/
Version History
- v1.0 (December 2025): Initial release
- PubMed integration via Biopython
- Plain-language summarization
- Automatic file saving
- Comprehensive templates
License
Part of the Vibe project. See main repository for license details.
Questions or Issues?
If you encounter problems or have suggestions for improvement, please open an issue in the Vibe repository.