convez.ai
MCP ServerIntegração & APIsAvançado
aks-builds

Healthcare Skills para Agentes de IA

aks-builds/healthcareskills

91 skills production-grade para IA construir software healthcare correto

Ver no GitHub
BomPluginScore · de 0 a 100

Conjunto de 91 skills de produção para engenharia de healthcare, cobrindo padrões FHIR, HL7v2, HIPAA, integração EHR e telehealth. Funciona como MCP server e garante conformidade clínica real, sem alucinações perigosas em dosagens ou dados médicos.

Instalação

$npx convez-ai aks-builds/healthcareskills

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
1
Acompanhando
1
Saúde
85%
Último commit
há 4 semanas
Versão
0.0.1
Issues abertas
8
Contribuidores
1

Casos de uso

  • Integração com sistemas EHR e standards FHIR/HL7
  • Validação de conformidade HIPAA e regulatória
  • Scaffolding de queries e processamento de dados clínicos

Veredito

MCP server robusto para domínio healthcare, com énfase em conformidade real e rejeição de alucinações perigosas. Excelente para equipes construindo software médico e precisando de domain knowledge hardcoded em padrões internacionais.

README

A tradução para o português deste README ainda está na fila.

🏥 healthcareskills

Give your AI agent the domain knowledge to build healthcare software correctly.

91 production-grade skills covering the full healthcare engineering stack — FHIR/HL7 interoperability, HIPAA compliance, clinical AI/ML, EHR integration, telehealth, revenue cycle, patient engagement, and more. Real-world clinical constraints built in. No hallucinated dosing. Works with Claude Code, Claude Desktop, Cursor, Zed, Windsurf, and any agent that speaks MCP or the Agent Skills spec.

Validate Skill Evals Skills: 91 MCP Agent Skills spec License: MIT

☝️ Claude Code invoking fhir_validate — one of 220+ typed MCP tools across healthcare domains.


Why healthcareskills

Healthcare software has a higher bar than most domains — HIPAA violations carry real fines, HL7 message errors delay patient care, and hallucinated drug dosing is dangerous. General-purpose AI agents don't know the difference between a FHIR R4 Patient.name.use value set and a free-text field. These skills do.

healthcareskills gives your agent accurate, up-to-date knowledge of FHIR R4/R5, HL7 v2, DICOM, HIPAA/HITECH, CDS Hooks, SMART on FHIR, ICD-10/SNOMED/LOINC, and 80+ more healthcare standards and regulations. The healthcare-context skill anchors every other skill to your specific org, patient population, EHR vendor, and regulatory jurisdiction — so advice is never generic.

Important: These skills are tools for software engineering in healthcare. They are not a substitute for clinical, legal, or regulatory advice, and they do not generate medical advice for patients. See AGENTS.md for safety rules.

How skills work together

healthcare-context is the foundation — every skill reads it first to understand your organization before offering guidance.

                        ┌──────────────────────────────────────┐
                        │          healthcare-context          │
                        │    (read by all skills first)        │
                        └──────────────────┬───────────────────┘
                                           │
   ┌──────────┬───────────┬────────────────┼────────────┬──────────┬──────────┐
   ▼          ▼           ▼                ▼            ▼          ▼          ▼
Clinical   Interop    Compliance       Digital      Revenue    Public &   Patient
Informatics           & Security       Health       Cycle      Pop.       Experience

See each skill's Related Skills section for the full dependency map.

Install

Option 1 — Claude Code Plugin (recommended)

/plugin marketplace add aks-builds/healthcareskills
/plugin install healthcare-skills

Option 2 — MCP Server (Claude Desktop, Cursor, Zed, Windsurf, custom agents)

Add to your mcp_servers config (e.g. .mcp.json or claude_desktop_config.json):

{
  "mcpServers": {
    "healthcare-skills": {
      "command": "npx",
      "args": ["github:aks-builds/healthcareskills/mcp"]
    }
  }
}

Or if you have it cloned locally:

{
  "mcpServers": {
    "healthcare-skills": {
      "command": "node",
      "args": ["/path/to/healthcareskills/mcp/index.js"]
    }
  }
}

Option 3 — Clone and Copy

git clone https://github.com/aks-builds/healthcareskills.git
cp -r healthcareskills/skills/* .agents/skills/

Option 4 — Git Submodule

git submodule add https://github.com/aks-builds/healthcareskills.git .agents/healthcareskills

Usage

Once installed, ask your agent naturally:

"Help me design a FHIR R4 integration for our patient portal"
→ fhir_validate + patient-portal + smart-on-fhir

"Do a HIPAA risk analysis on this architecture diagram"
→ hipaa_checklist + healthcare-cybersecurity + audit-logging

"Set up an EDI 837 claims submission flow"
→ billing-claims + medical-coding

"Build a CDS Hook that warns on drug-drug interactions"
→ clinical-decision-support + medication-reconciliation + terminology-services

"Is our ambient scribe compliant with HIPAA and FDA SaMD rules?"
→ ambient-clinical-intelligence + fda-samd + hipaa_checklist

Or invoke skills directly:

/fhir-integration
/hipaa-compliance
/ehr-integration
/clinical-ai-ml

MCP tool calls (any MCP host):

fhir_validate({ resource: "Patient", payload: { ... } })
hipaa_checklist({ component: "audit-logging", context: "patient portal" })
hl7_troubleshoot({ message_type: "ADT^A01", error: "MSH segment missing" })

Available Skills

Skill Categories

Foundation

  • healthcare-context — Org/patient/jurisdiction/EHR context — every other skill reads this first

Clinical Informatics & EHR

  • ehr-integration · clinical-decision-support · clinical-documentation · cpoe-orders · smart-on-fhir · medication-reconciliation

Interoperability

  • fhir-integration · hl7-v2 · dicom-imaging · cda-ccda · ihe-profiles · terminology-services · tefca-hie
  • New: fhir-shorthand · fhir-subscriptions · fhir-bulk-data · ncpdp-pharmacy · smart-health-cards · omop-cdm · direct-trust-messaging · fhirpath-cql

Compliance & Security

  • hipaa-compliance · phi-handling · healthcare-cybersecurity · 21-cfr-part-11 · gdpr-health-data · hitrust-csf · audit-logging
  • New: cms-interoperability-rules · information-blocking · ehr-certification-uscdi · 42-cfr-part-2 · nist-csf-healthcare · soc2-healthcare · fedramp-healthcare

Digital Health & AI/ML

  • telehealth-platform · patient-portal · remote-patient-monitoring · wearables-integration · clinical-ai-ml · fda-samd · health-chatbots
  • New: digital-therapeutics · mental-health-tech · care-coordination · social-determinants · care-plan-management · chronic-disease-management · ai-governance-healthcare · ambient-clinical-intelligence · nlp-clinical-text · digital-pathology · predictive-analytics-clinical · clinical-llm-safety

Revenue Cycle & Operations

  • medical-coding · billing-claims · prior-authorization · value-based-care
  • New: revenue-cycle-analytics · credentialing · remittance-processing · population-health-programs

Public & Population Health

  • public-health-reporting · population-health-analytics · clinical-research

Patient Experience

  • health-content-writing · patient-engagement · accessibility-healthcare
  • New: patient-matching-mpi · consent-management · health-equity-analytics

Cloud & Engineering

  • New: healthcare-cloud-aws · healthcare-cloud-azure · healthcare-cloud-gcp · healthcare-devops · healthcare-observability · healthcare-microservices

Data & Analytics

  • health-data-lake
  • New: clinical-data-warehouse · real-world-evidence · synthetic-patient-data · fhir-analytics · healthcare-data-quality

Specialized / Cross-cutting

  • medical-imaging-ai · genomics-precision-medicine
  • New: oncology-informatics · pharmacy-informatics · laboratory-informatics · behavioral-health · genomics-data-standards · decentralized-clinical-trials · healthcare-iot · voice-healthcare · ar-vr-healthcare

Repository layout

healthcareskills/
├── .claude-plugin/
│   ├── plugin.json            Claude Code plugin manifest
│   └── marketplace.json       Claude plugin marketplace entry
├── mcp/
│   ├── index.js               MCP server entrypoint
│   ├── loader.js              SKILL.md parser + skill-graph resolver
│   └── manifest.js            Generates tool manifest from all skills
├── skills/
│   └── skill-name/
│       ├── SKILL.md           Skill instructions (v2 frontmatter)
│       ├── evals/
│       │   └── evals.json     5–6 eval scenarios (CI-benchmarked)
│       └── references/        Optional deep-dive docs loaded on demand
├── AGENTS.md                  Guidelines for AI agents working in this repo
├── CLAUDE.md                  Claude Code reminders
├── CONTRIBUTING.md
├── LICENSE
├── README.md
├── SECURITY.md
└── VERSIONS.md

Contributing

PRs and issues welcome. See CONTRIBUTING.md — note the elevated bar for safety and accuracy in healthcare content. Clinical claims, regulatory references, and standard field values must be verified before merging.

License

MIT — use these however you want, but verify before relying on them in safety-critical clinical systems.

Ver o original no GitHub

Bom para

Relacionados