Top 8 OpenClaw Skills for Healthcare Clinical Documentation

1. clinical-reports

The most comprehensive documentation skill in the library. clinical-reports generates case reports following CARE guidelines, diagnostic reports for radiology, pathology, and lab results, clinical trial reports per ICH-E3 standards, and patient encounter documentation including SOAP notes, history and physical exams, and discharge summaries.

Key strengths:

• Covers the full range of clinical document types in a single skill module
• Includes a HIPAA de-identification checklist covering all 18 identifiers defined under the Safe Harbor method
• Requires the companion scientific-schematics skill for AI-generated figures like CONSORT diagrams and patient timelines, enforcing visual documentation standards
• References FDA 21 CFR Part 11 for electronic records compliance and ICH-GCP guidelines for clinical trials

Limitations:

• No direct API connections to EHR systems. The skill generates standalone documents but does not push them into Epic, Cerner, or other platforms.
• The mandatory scientific-schematics dependency adds installation complexity. You need both skills for the full workflow.

Best for: Clinical teams that need structured, publication-ready or chart-ready documentation across multiple report types from a single skill installation.

2. clinical-decision-support

Generates population-level clinical decision support documents with GRADE evidence grading. This skill focuses on treatment recommendation reports, cohort analyses with biomarker stratification, and evidence synthesis for pharmaceutical and research settings.

Key strengths:

• Implements the full GRADE evidence grading system from 1A (strong recommendation, high-quality evidence) through 2C (weak recommendation, low-quality evidence)
• Supports genomic biomarker integration including mutations, copy number variations, and gene fusions
• Produces publication-ready LaTeX and PDF output with statistical tables and TikZ flowcharts
• Integrates bidirectionally with the clinical-reports and treatment-plans skills, so population-level insights can feed individual treatment documentation

Limitations:

• Designed for population-level analysis, not individual patient care decisions at the point of care
• No direct connections to external clinical databases. All evidence must be provided as input to the agent.

Best for: Research teams building treatment guidelines, pharmaceutical companies developing evidence dossiers, and clinical informaticists creating decision algorithms for formulary committees.

3. prior-auth-review-skill

Automates payer review of prior authorization requests through a two-phase workflow. Phase one handles intake: collecting PA details, validating provider credentials and diagnosis codes, searching applicable coverage policies, and extracting clinical evidence from supporting documents. Phase two generates the authorization decision and provider notification letter.

Key strengths:

• Connects to four real APIs: NPI for provider credential verification, ICD-10 for diagnosis code validation, CMS Coverage for Medicare policy lookup (NCDs and LCDs), and CMS Fee Schedule for CPT/HCPCS procedure code validation
• Generates structured outputs including assessment JSON, decision JSON, and provider notification letters with authorization numbers or detailed denial rationale with appeal information
• Targets automatic approval for 40% to 60% of straightforward cases according to the skill documentation
• Accepts clinical notes, H&P documents, diagnostic imaging reports, and lab results as supporting inputs

Limitations:

• Built specifically for payer organizations, not provider-side pre-authorization submission workflows
• The claimed reduction in review time (from 30 to 60 minutes down to under 5 minutes) comes from the skill documentation and has not been independently validated

Best for: Health insurance payers processing high volumes of prior authorization requests, including Medicare Advantage plans, commercial insurers, and Medicaid managed care organizations.

4. fhir-developer-skill

A development-focused skill for building FHIR R4-compliant REST endpoints. Rather than generating clinical documents directly, this skill teaches an OpenClaw agent how to construct healthcare APIs that other documentation systems can connect to.

It covers seven core FHIR resources: Patient, Observation, Encounter, Condition, MedicationRequest, Medication, and Bundle.

Key strengths:

• Implements SMART on FHIR OAuth authorization with scope enforcement for secure access control
• Covers CRUD operations, search pagination, conditional operations, and transaction/batch processing
• References four major clinical terminology systems: LOINC, SNOMED CT, RxNorm, and ICD-10
• Includes proper error handling with OperationOutcome-formatted responses and ETag-based version control

Limitations:

• A development guide, not a documentation generator. It helps agents build FHIR infrastructure rather than producing clinical notes.
• Covers only seven of the 150+ FHIR resources defined in the R4 specification

Best for: Developer teams building healthcare integrations that need FHIR-compliant APIs, particularly for connecting clinical documentation systems to EHR platforms or health information exchanges.

5. clinical-trial-protocol-skill

An orchestrator skill that generates complete clinical trial protocols for medical devices and drugs. It runs in two modes: research-only (gathering intervention data and surveying existing trials for a formatted summary) or full protocol mode (producing a complete 12-section protocol document from synopsis through statistical analysis plan).

Key strengths:

• Connects to ClinicalTrials.gov via MCP server for searching existing trials by condition, intervention, sponsor, location, status, and phase
• References FDA regulatory pathways including 510(k), PMA, De Novo, IND, NDA, and BLA, helping teams identify the correct submission route
• Includes automated sample size calculations using scipy and numpy with appropriate statistical tests
• Generates intervention metadata, clinical research summaries, protocol foundation sections, and operations documentation as separate artifacts

Limitations:

• Explicitly requires biostatistician review, FDA Pre-Submission meetings, and IRB approval before clinical use. The outputs are drafts, not submission-ready protocols.
• Protocol structure follows U.S. regulatory conventions. Teams submitting to EMA, PMDA, or other non-FDA regulators will need to adapt the output format.

Best for: Medical device and pharmaceutical companies drafting initial protocol documents for FDA submission, particularly during feasibility studies and pilot trial design.

6. lit-synthesizer

Automates biomedical literature discovery, analysis, and synthesis. The lit-synthesizer queries PubMed via the NCBI Entrez API using MeSH terminology, searches bioRxiv and medRxiv for recent preprints, and generates structured literature reviews with integrated citations.

Key strengths:

• Direct PubMed access with MeSH terminology support for precise biomedical searches
• Searches preprint servers for unpublished research, catching findings before formal peer review
• Produces ranked paper lists, LLM-generated summaries, and drafted review paragraphs with citations ready for manuscript integration
• Includes research gap detection based on keyword distribution analysis, highlighting underexplored areas in the literature

Limitations:

• Full-text access depends on open-access availability. The skill retrieves abstracts from PubMed but cannot access paywalled full-text articles.
• Citation relationship visualization and network mapping features are still in active development.

Best for: Researchers preparing systematic reviews, grant writers assembling background literature sections, and clinical teams monitoring the evidence base for specific conditions or interventions.

7. health-trend-analyzer

Tracks and correlates health data patterns over time across multiple dimensions: weight and BMI trajectories, symptom frequency and recurrence, medication adherence, laboratory result trends, and mood and sleep quality scores.

Key strengths:

• Runs entirely locally with no external API calls, keeping all patient health data private on the machine where the agent operates
• Produces both text-based trend reports and interactive HTML dashboards with six chart types powered by ECharts
• Includes change detection that flags significant variations, deterioration patterns, improvements, and threshold warnings for metrics like BMI extremes
• Correlates medications with symptom changes and lifestyle factors with emotional states, surfacing patterns that manual chart review might miss

Limitations:

• Designed for individual patient health tracking, not population-level epidemiological analysis or cohort studies
• Requires structured JSON input files covering profile data, symptom logs, and test results. Unstructured clinical notes need preprocessing before the skill can analyze them.

Best for: Patient wellness monitoring programs, chronic disease management teams, and health coaching platforms that need to track and visualize outcomes over time.

8. medical-imaging-review

Generates comprehensive literature reviews for medical imaging AI research, covering deep learning applications in segmentation, detection, and classification across CT, MRI, X-ray, ultrasound, and digital pathology modalities.

Key strengths:

• Connects to ArXiv for preprints, PubMed for clinical validation studies, and Zotero for reference library management across three independent retrieval channels
• Follows a structured 7-phase manuscript generation workflow from topic scoping through final manuscript assembly
• Includes comparison tables with quantitative performance metrics like Dice scores and Hausdorff distance for direct method comparison
• Enforces appropriate hedging language ("may," "suggests," "appears to") and requires citation support for every factual claim in the generated review

Limitations:

• Produces literature reviews about imaging AI methods, not diagnostic radiology reports. It does not analyze actual medical images.
• Targets academic manuscripts with 80 to 120 references, which is substantially more depth than most clinical documentation teams need.

Best for: Academic researchers and radiology informatics teams writing survey papers, background sections for grant applications, or systematic reviews of imaging AI performance across modalities.