Skill UI

This skill guides users in analyzing complex questions related to criminal or legal incidents (e.g., murder, theft, arson, accident). By first classifying the specific case type and identifying domain-specific keywords, it dramatically improves the precision of subsequent search queries. It teaches users how to combine case type terms with distinctive constraints like amounts, dates, and locations for maximum search effectiveness.

This skill simulates the rigorous peer review process used in top-tier AI and machine learning conferences (like NeurIPS). It helps users diagnose potential weaknesses in a submission regarding quality, clarity, significance, ethical compliance, and reproducibility. It guides the identification of decision-relevant issues that ACs and reviewers are likely to raise, helping authors strategically plan responses for revision.

This skill audits OpenMed NER or de-identification models to detect performance disparities across sensitive demographic subgroups (e.g., sex, age, race). It calculates metrics like per-group recall and leakage disparity, helping users identify instances of 'under-protection' before deployment, ensuring equitable AI governance and compliance.

This skill automates the creation of comprehensive Model Cards for OpenMed clinical NER or de-identification models. It gathers data from various OpenMed evaluation reports (e.g., `GateReport`, `fairness_report`, `error_report`) to structure critical governance information, including intended use, quantitative performance metrics, subgroup analysis, limitations, and required medical-device disclaimers. It ensures documentation is reproducible and compliant for clinical AI deployment.

A robust, on-device solution for processing large-scale clinical text corpora. Supports batch NER, PII extraction, and de-identification over thousands of notes. Features include sharding, checkpointing, resumable functionality, and streaming results to JSONL, ensuring PHI safety and constant memory usage for massive datasets.

This skill deploys the OpenMed Model Context Protocol (MCP) server, exposing critical clinical NLP tools such as NER, PII extraction, and de-identification. It enables AI coding agents and chat clients to interact with OpenMed's capabilities via simple tool calls, running locally on-device to ensure user data privacy and security. Supports both stdio and Streamable HTTP deployment modes.

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.

This skill implements a critical CI gate to ensure the privacy and performance integrity of de-identification models. It automatically fails the build if the model's recall on held-out PHI data drops below the defined threshold, or if any critical Personally Identifiable Information (PHI) is leaked (leakage must be zero). It enforces strict privacy standards in continuous integration pipelines, blocking unsafe merges.

This module converts diverse clinical inputs—including scanned images, faxes, CSV/TSV exports, and C-CDA XML—into a normalized, clean text format (`ExtractedDocument`). It supports on-device OCR (Tesseract/PaddleOCR) and structured redaction for tables. It is the crucial first stage for feeding de-identification and Named Entity Recognition (NER) pipelines.

This technique segments unstructured clinical documents (such as HPI, PMH, and A&P sections) into contextually rich, canonical chunks. This is crucial for improving the precision of downstream NLP tasks like Named Entity Recognition (NER) and de-identification. By identifying and normalizing headers and mapping them to LOINC codes, the system ensures that the source context (e.g., historical vs. active) is maintained, leading to significantly more accurate results.

Provides a hardened FastAPI endpoint for OpenMed's core functionalities: clinical Named Entity Recognition (NER), Personally Identifiable Information (PII) extraction, and de-identification. Designed for self-hosted microservices, it handles inference requests, model lifecycle management, and supports dynamic batching, ensuring secure, on-device processing of sensitive clinical data.

This skill applies consistent date shifting to clinical text, ensuring that absolute dates are obscured while critically preserving the temporal intervals between medical events. It is essential for longitudinal research or survival analysis where time-to-event data must be preserved under HIPAA Safe Harbor rules. It maintains the integrity of time differences (e.g., length of stay, visit gaps) by applying a stable, per-patient offset.