Agent Skills for Scientific Research: From Literature Review to Drug Discovery

Scientific research requires stitching together dozens of tools: databases for querying, Python packages for analysis, visualization libraries for figures, and document tools for reports. The k-dense-ai/claude-scientific-skills collection gives your AI agent working knowledge of 140 scientific skills across 16 domains, so you can describe a research pipeline in plain English and have the agent execute it.

What Is Inside

This is not a general-purpose skill set. It is purpose-built for scientific workflows:

Each skill includes documentation, code examples, use cases, and integration guides. They follow the open Agent Skills standard (SKILL.md format).

Installation

# Clone the repo
git clone https://github.com/k-dense-ai/claude-scientific-skills.git

# Install for Claude Code (global)
cp -r claude-scientific-skills/scientific-skills/* ~/.claude/skills/

# Or install for a specific project
mkdir -p .claude/skills
cp -r /path/to/claude-scientific-skills/scientific-skills/* .claude/skills/

Restart your agent after installation. Skills activate automatically based on the research task.

Real Workflow: Drug Discovery Pipeline

You want to find potential EGFR inhibitors for lung cancer. This pipeline chains 8 skills together.

Open a project directory in Claude Code:

mkdir -p ~/egfr-research && cd ~/egfr-research
claude

Run the full pipeline:

Use available skills to: query ChEMBL for EGFR inhibitors with IC50 below 50nM,
analyze structure-activity relationships with RDKit, generate improved analogs,
perform virtual screening against AlphaFold EGFR structure, search PubMed for
resistance mechanisms, and create a comprehensive report.

The agent chains the skills automatically:

1. ChEMBL queries bioactivity data for EGFR inhibitors
2. RDKit computes molecular descriptors and identifies SAR patterns
3. datamol generates structurally similar analogs
4. DiffDock docks candidates against the AlphaFold EGFR structure
5. PubMed searches for known resistance mechanisms
6. Scientific Visualization creates publication-quality figures

Each step uses real APIs and real data. The output is a structured research report with visualizations.

Real Workflow: Literature Review with PubMed and OpenAlex

You need a literature review on a specific topic.

Search PubMed and OpenAlex for publications on CRISPR-based gene therapy for
sickle cell disease from the last 2 years. Extract key findings, group by
theme, and create a structured review with proper citations.

The agent uses:

1. PubMed for biomedical literature search
2. OpenAlex for broader academic coverage
3. Citation Management for formatting references
4. Scientific Writing for structuring the review

Real Workflow: Single-Cell RNA-seq Analysis

Load the 10X dataset from data/pbmc.h5ad, perform QC and doublet removal,
identify cell types using known markers, run differential expression, enrich
pathways via KEGG, and visualize the results.

The agent uses Scanpy for analysis, NCBI Gene for markers, KEGG for pathways, and Plotly for interactive visualizations.

Tips

• Prefix prompts with "Use available skills" to encourage the agent to chain skills rather than improvising.
• Install uv (curl -LsSf https://astral.sh/uv/install.sh | sh) for automatic Python dependency management.
• Keep datasets in your project directory. The skills work best when data files are local.
• For clinical research, combine ClinVar (variant pathogenicity) with COSMIC (cancer mutations) for comprehensive variant interpretation.

When Not to Use This

Skip this if you are not doing scientific research. The skills are specialized for bioinformatics, chemistry, clinical research, and data analysis. For general software development, use addyosmani/agent-skills or the official Anthropic skills.

Explore more specialized skills on the SkillMap leaderboard.