Skill UI

This tool systematically audits and scores defined AI agents to ensure they are correctly scoped, highly specific, and safe for programmatic orchestration. It checks for description ambiguity, over-permissive tool grants, and anti-patterns that break headless workflows. Use it when onboarding new agents, integrating agent fleets, or debugging inconsistent multi-agent selection.

This skill evaluates de-identification and clinical NER models against rigorous leakage-first release gates (G1a-G8). Instead of relying on aggregate F1 scores, it focuses on detecting residual Protected Health Information (PHI) leakage and enforcing per-label recall floors, which is crucial for HIPAA compliance. Determines if a model is RELEASABLE or QUARANTINED.

Utilizes OpenMed's analyze_text to perform sophisticated Named-Entity Recognition (NER) on clinical and biomedical text. This tool extracts structured entities such as diseases, drugs, anatomy, genes, and proteins. It provides precise character spans and model confidence scores, making it essential for building robust, privacy-first clinical Natural Language Processing (NLP) pipelines.

This skill maps free-text laboratory and clinical observation names to standardized LOINC codes. It resolves ambiguous terms by considering the full six-axis model: Component, Property, Time, System (specimen), Scale, and Method. It integrates with FHIR terminology services to ensure accuracy, supporting US Core Laboratory Results and UCUM units.

This skill standardizes free-text medication mentions extracted by OpenMed to RxNorm RxCUIs using the public RxNav API. It resolves complex relationships between brand names, generic ingredients, strengths, and dose forms, providing a stable, canonical code. Ideal for clinical interoperability, data analytics, de-duplication, and building US Core Medication resources.

This skill facilitates the crucial grounding step for clinical coding by interacting with external FHIR terminology servers. It supports core operations such as `$validate-code`, `$expand` (for ValueSets), `$lookup`, and `$translate` (ConceptMaps). It ensures that raw text spans are converted into validated, coded `CodeableConcept` resources using industry standards like SNOMED CT, LOINC, and ICD-10, maintaining interoperability.

This skill guides users on deploying OpenMed models completely on-device for maximum privacy and edge computing. It details running workflows using three backends: MLX (Apple Silicon), CoreML (iOS/macOS), and ONNX/WebGPU (cross-platform/browser). The process involves converting, optionally quantizing, and running models locally, ensuring PHI never leaves the host device.

This tool simulates a senior application security engineer, performing a comprehensive, multi-phase audit. It checks for exposed secrets, prompt injection vectors, dependency vulnerabilities (supporting multiple languages), and critical configuration misconfigurations. It provides a scored, prioritized remediation report, ensuring a robust security posture before deployment.

This skill generates a comprehensive, structured catalogue of all key concepts found within a talk's summary and timeline. It categorizes each concept and scores its potential for presentation (HIGH/MEDIUM/LOW) to help users identify the strongest narrative angles. The output is designed to serve as a structured inventory, guiding the subsequent positioning and content development stages.

This skill validates FHIR R4 resources and Bundles against mandatory US Core profiles (USCDI). It goes beyond mere syntactic checks to ensure that data prepared for submission to major Electronic Health Records (EHRs) like Epic or Cerner meets the required US-specific conformance standards. It detects common data gaps, checks 'must-support' elements, and converts the comprehensive validation report into a standard FHIR OperationOutcome for seamless pipeline integration.

This skill processes verbose, stream-of-consciousness voice dictations (from tools like macOS Dictation or Wispr Flow) and transforms them into highly structured, token-efficient prompts. It removes filler words, repetitions, and tangents, organizing the core intent into standard sections (Context, Objective, Constraints) suitable for advanced LLMs like Claude Code.