
databricks-industry-solutions/vital-skills
Developer toolsHealth & Life Sciences agent skills for Databricks Genie Code: genomics, RWE cohorts, PHI de-identification, payer/provider measures, and skill evaluation
Overview
Each skill is a SKILL.md folder that teaches Genie Code following the Agent Skills standard) how to run domain workflows with libraries, tools and MCP servers. Site: https://databricks-industry-solutions.github.io/vital-skills/ Git clone this repo onto Databricks. Then open Genie Code and click on Customizations to add the cloned folder. It should point to the skills subfolder Run sync_skills_git2unity.py to sync the latest skills from the repo to Unity Catalog. They also show up on Unity Gateway under Skills. On Databricks, start Serverless or a cluster (>= 15.0 runtime) for compute. Then open the terminal (how-to-guide). It should already have the Databricks CLI installed. You can get the SQL warehouse id from the Databricks left menu bar: SQL Warehouses > select warehouse > Name. It should list the warehouse ID. More details here. Repo layout and skill templates are inspired by the patterns in SciAgent-Skills Licensed under the Databricks License. See LICENSE.md and NOTICE.
README
Vital Skills
Agent skills for Health & Life Sciences workflows.
Each skill is a SKILL.md folder that teaches Genie Code following the Agent Skills standard) how to run domain workflows with libraries, tools and MCP servers.
Site: https://databricks-industry-solutions.github.io/vital-skills/
Setup
Git clone this repo to Databricks
Git clone this repo onto Databricks. Then open Genie Code and click on Customizations to add the cloned folder. It should point to the skills subfolder
Available Skills
| Skill | Category | What it does |
|---|---|---|
| bulk-rnaseq | Bioinformatics | Find top differentially expressed genes from bulk RNA-seq counts. |
| open-weight-models | Bioinformatics | Package, register, validate & deploy open-weight HLS models on Databricks, such as TEDDY and Geneformer. |
| pathway-enrichment-analysis | Bioinformatics | Pathway enrichment for RNA-seq (ORA or GSEA). |
| sc-rnaseq | Bioinformatics | Analyze single-cell RNA-seq data from .h5ad files. |
| cohort-builder | Clinical | Build a reproducible, feasibility-checked patient cohort from coded + free-text clinical data. |
| phi-deidentifier | Clinical | HIPAA Safe-Harbor de-identify a Unity Catalog table (k-anonymity + governed view). |
| rwe-cohortstudy | Clinical | Comparative effectiveness (cohort study) with propensity adjustment (matching, IPW). |
| semantic-builder | Clinical | Canonical payer+provider measure catalog mapped to Unity Catalog metric views. |
| accelerators | Others | Route imaging, biology-model and HL7v2 tasks to an HLS accelerator (Pixels, Genesis Workbench or funke-hl7v2) with the exact next step. |
| presentation-builder | Others | Turn a cohort, dataset, model, or analysis into a grounded, shareable HTML deck (PHI de-id delegated to phi-deidentifier). |
| report-builder | Others | LaTeX report generation and PDF compilation from analysis notebooks via Tectonic. |
| scientific-figures | Others | Publication-quality scientific and biopharma figures for notebooks, apps, and reports. |
| skill-eval | Others | Benchmark a skill: paired runs with/without it, MLflow genai.evaluate scoring, win/loss + failure taxonomy. |
Repository Layout
vital-skills/
├── AGENTS.md # Skill authoring guide
├── CLAUDE.md # Compatibility shim → AGENTS.md
├── templates/ # Pipeline / toolkit / guide templates
└── skills/
├── bulk-rnaseq/
├── cohort-builder/
└── …
Each skill:
skills//
├── SKILL.md # Required
├── references/ # Optional — loaded on demand
├── assets/ # Optional
└── scripts/ # Optional
Creating a Skill
See CONTRIBUTING.md.
Repo layout and skill templates are inspired by the patterns in SciAgent-Skills
License
Licensed under the Databricks License. See LICENSE.md and NOTICE.md.
Disclaimer
This is an open source solution to provide agent skills to assist with healthcare and life sciences tasks. However, all compliance obligations remain with the user. This includes but is not limited to:
- HIPAA: You are responsible for ensuring your deployment meets HIPAA requirements (encryption, access controls, audit logging, BAAs, etc.)
- GDPR, CCPA, and other privacy regulations: Evaluate whether your use of this tool satisfies applicable data protection laws
- Validation: You must verify that results are complete and accurate for your specific data and use case
- Data Encryption: Enable encryption at rest and in transit in your Databricks workspace
- Access Controls: Configure appropriate table/catalog permissions in Unity Catalog
- Audit Logging: Enable workspace audit logs for compliance tracking
Databricks makes no guarantees that use of this solution alone is sufficient for regulatory compliance
Contributors
- Yen Low
- Zachary Phillips
- May Merkle-Tan
- Praneeth Paikray
- Vimal Thomas Joseph
- Peter Hawkins
- Rohan Parikh
To add yourself, see CONTRIBUTING.md.
Acknowledgements
Thanks to Eli Swanson, Douglas Moore, Parastou Eslami and Khagay Nagdimov for their advice.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add databricks-industry-solutions/vital-skills