550+ prompt templates for medical research agents

What it does

This is a curated collection of 550+ prompt-style “skills” for AI agents (Claude Code, OpenClaw, Hermes Agent, etc.) that cover the full medical research workflow: designing search strategies, planning clinical protocols, running bioinformatics analyses, and drafting manuscripts. Each skill is essentially a structured prompt template with embedded medical research logic, not executable code.

The interesting bit

The granularity is almost obsessive. Instead of one generic “analyze my data” skill, you get Elastic Net Feature Selection, Batch Effect Correction, Figure-First Paper Reader, and Contradictory Findings Resolver as separate, purpose-built modules. There’s even a MedSkillAudit sub-project for auditing the skills themselves — meta enough to notice.

Key highlights

• 554 skills across Evidence Insights, Protocol Design, Data Analysis, Academic Writing, and general tasks
• Explicitly designed for omics-heavy work: single-cell, bulk RNA-seq, spatial transcriptomics, multi-omics integration
• Includes niche but practical tools like Preprint Surveillance Finder, Population Gap Detector, and Medical English Precision Editor
• Awesome Med Research Skills sub-collection highlights 140+ “high-quality” skills with more professional logic
• All skills are claimed to be originally developed in-house, not scraped or crowdsourced

Caveats

• The README is long on skill listings and short on how the prompts actually perform — no evaluation metrics or comparison studies shown
• “Works with” compatibility claims (OpenClaw, Opencode) are badge-level; no technical detail on skill format or integration mechanism
• The star-growth chart is included as a candidate image, which tells you something about priorities

Verdict

Medical researchers already using AI agents for repetitive tasks will find this a useful prompt pantry, especially for bioinformatics and clinical protocol work. If you’re looking for executable pipelines or verified clinical decision support, this is explicitly not that — it’s structured prompting, not code.