# Route medical and biomedical tasks with OpenClaw

Medical and biomedical tasks require navigating hundreds of specialized capabilities. This skill routes Codex to the right OpenClaw capability for clinical, genomics, drug discovery, bioinformatics, and regulatory workflows. It enforces safety boundaries so outputs support, never replace, qualified clinicians.

## Install

```bash
npx skillstore add yuan1z0825/openclaw-medical-skills
```

## Metadata

- - Slug: yuan1z0825-openclaw-medical-skills
- - Version: 1.0.0
- - Author: Yuan1z0825
- - GitHub username: Yuan1z0825
- - License: MIT
- - Repository: https://github.com/Yuan1z0825/nature-skills/tree/main/skills/openclaw-medical-skills
- - Ref: main
- - Supported tools: Claude, Codex, Claude Code
- - Risk level: low
- - Risk factors: external\_commands, network, filesystem
- - Quality score: 73
- - Public page: https://skillstore.pages.dev/skills/yuan1z0825-openclaw-medical-skills
- - Manifest: https://skillstore.pages.dev/api/skills/yuan1z0825-openclaw-medical-skills/manifest

## Capabilities

- Routes Codex to 869 OpenClaw medical capabilities across 8 categories: general, clinical, scientific databases, bioinformatics, omics, ClawBio pipelines, BioOS extensions, and data science.
- Searches capability index by skill slug, category, or task description to identify the right tool for a request.
- Classifies requests by risk and domain \(patient care, clinical research, literature review, drug/variant interpretation, bioinformatics, regulatory\).
- Provides output standards by task: plain-language clinical summaries, evidence-graded literature reviews, drug research with mechanism and monitoring, and ACMG-compliant variant interpretation.
- Enforces safety boundaries: outputs are for education, research, drafting, and decision support only, never definitive diagnosis or prescription.
- Recommends primary sources \(PubMed, ClinicalTrials.gov, FDA/EMA labels, ClinVar, gnomAD, COSMIC, OMIM, Orphanet, HPO\) over fabricated data.

## Use Cases

- Clinical Literature Review for a Clinician: A clinician needs an evidence summary on a drug interaction for a complex patient. The skill routes to PubMed search and drug-drug interaction analysis, returns graded evidence with source dates and citations.
- Genomic Variant Interpretation for a Researcher: A genomics researcher has a VCF file and needs ACMG-classified variant interpretation. The skill routes to variant analysis tools, integrates ClinVar and gnomAD annotations, and produces a decision-support draft for qualified genetics review.
- Drug Discovery Pipeline for a Biotech Analyst: A biotech analyst needs target intelligence, ADMET profile, and repurposing candidates for a compound. The skill routes to drug research, chemical safety, and network pharmacology capabilities, returning structured reports with primary-source citations.

## Prompt Templates

### Beginner: Find the right capability

```
I have a task involving [clinical topic / dataset / drug name / gene]. Search the OpenClaw capability index and tell me which 2-3 capabilities best match, with a one-line summary of each.
```

### Intermediate: Classify and route a research request

```
Classify this request by risk and domain, then route it to the best OpenClaw capability: [paste your biomedical or clinical request]. State the chosen capability, why it fits, and what primary sources you will use.
```

### Advanced: Multi-capability drug interaction analysis

```
Using the drug-drug interaction and pharmacovigilance capabilities, analyze the interaction between [Drug A] and [Drug B] for a patient on [Drug C]. Include CYP450 mechanism, severity, monitoring, alternatives, and escalation criteria. Cite the label and guideline sources with retrieval dates.
```

### Advanced: ACMG variant interpretation workflow

```
Walk me through ACMG/AMP variant interpretation for [gene] [variant, e.g., BRCA1 c.5266dupC]. Route to variant interpretation and cancer variant interpretation capabilities, integrate ClinVar/gnomAD/COSMIC, and produce a draft classification for qualified genetics review. Mark all outputs as decision support, not diagnosis.
```

## Limitations

- Does not include the actual OpenClaw or NanoClaw runtime; assumes Codex-native tools, local files, Python/R, and web access.
- Some capabilities require credentials, MCP endpoints, or local databases not available in every Codex session; skill will state the limitation rather than fabricate results.
- Not a substitute for qualified clinicians, pharmacists, genetic counselors, or regulatory professionals.
- Capability index is a routing reference, not an executable skill set; downstream execution depends on the Codex environment.

## Best Practices

- Always state the chosen capability and primary sources before producing a draft; transparency beats silent routing.
- De-identify patient information \(names, dates, identifiers\) before using clinical examples; keep analyses local unless the user asks otherwise.
- Mark every clinical or regulatory output as a draft requiring qualified professional review; never present as definitive diagnosis or clearance.
- Cite retrieval dates for time-sensitive facts \(FDA labels, trial status, guidelines\) so users know how current the evidence is.

## Anti Patterns

- Do not present the skill's output as a definitive diagnosis, prescription, treatment order, or substitute for a qualified clinician.
- Do not fabricate patient data, lab values, guideline recommendations, regulatory status, trial availability, or database results when the source is unavailable.
- Do not skip the safety boundary or crisis routing: emergencies, self-harm risk, and severe acute symptoms must trigger immediate emergency-services guidance.

## Security Audit

- - Safe to publish: true
- - Audited at: 2026-06-24T06:12:47.976\+00:00
- - Summary: Static analysis flagged 234 potential issues across 4 files \(3,657 lines\), but evaluation confirms nearly all are false positives. The 'Weak cryptographic algorithm' findings \(e.g., MD5/SHA\) are matching the medical acronyms 'MD' \(Medical Doctor\) and 'AMP' \(ACMG/AMP, AMP/ASCO/CAP\) used throughout the capability index. The 'Windows SAM database' findings are matching the bioinformatics file format 'SAM/BAM' \(Sequence Alignment Map\). 'Ruby/shell backtick execution' findings are matching legitimate markdown inline code like \`bcftools consensus\`. The 'Hidden file in home directory' findings are matching documented installation paths like \`~/.openclaw/skills/\`. No actual code execution, exfiltration, or credential access patterns were found. The skill is a routing/reference document with safety boundaries; no executable scripts exist.

## Stats

- - Views: 0
- - Downloads: 1
- - Favorites: 0
- - Popularity score: 0