Skill UI

Builds NeMo Guardrails-based input/output validation pipelines to block prompt injection, redact PII, enforce content policies, and verify structured outputs so LLM-powered services stay safe before downstream use.

Run Hugging Face models directly in JavaScript/TypeScript for text, vision, and speech tasks without backend servers, using Transformers.js in Node.js or modern browsers with WebGPU, quantization, and pipeline abstractions for efficient inference.

Transforms user articles, reports, or documentation into natural two-host conversational podcasts by building structured JSON scripts and running the bundled TTS pipeline to emit MP3 audio plus transcripts.

Provides patterns and techniques for building robust evaluation loops for AI agents. This skill guides the implementation of self-critique, evaluator-optimizer pipelines, and test-driven refinement workflows. Use it to significantly enhance the accuracy and quality of complex outputs like code, reports, and analyses by implementing structured, iterative improvement cycles (Generate → Evaluate → Critique → Refine).

Three-layer verification pipeline that extracts verifiable claims from AI output, searches the web for supporting or contradicting sources, flags hallucinations, and builds structured reports for human review.

PyHealth is a modular Python toolkit designed for building deep learning pipelines using complex medical data, including EHR records, physiological signals, and medical imaging. It provides a standardized 5-stage workflow: Dataset → Task → Model → Trainer → Metrics. This allows users to train sophisticated models for clinical prediction (e.g., mortality, readmission, drug recommendation) while handling medical code mapping (ICD, ATC, RxNorm) and ensuring proper data splitting for rigorous research.

A comprehensive skill for leveraging massive pre-trained transformer models across diverse AI domains, including Natural Language Processing (NLP), Computer Vision (CV), and Audio. Use this skill for advanced tasks such as text generation, classification, question answering, summarization, object detection, and model fine-tuning on proprietary datasets. It provides streamlined pipelines and granular control for deployment.

A comprehensive guide detailing five robust architectural patterns for building applications powered by Claude, including Chatbot, RAG, Agent, Content Pipeline, and Evaluation. It provides complete code examples and decision matrices to help developers select the optimal architecture based on project requirements, cost, and complexity.

Automates comprehensive regression testing for all enabled cloud Large Language Model (LLM) providers. This includes verifying core connectivity, standard chat completion, structured JSON output, and Server-Sent Events (SSE) streaming capabilities. Essential for ensuring the reliability and stability of multi-cloud AI integrations in a CI/CD pipeline.

This skill establishes a comprehensive CI/CD pipeline specifically for Abridge clinical AI integrations using GitHub Actions. It enforces critical healthcare compliance steps, including FHIR R4 resource validation, robust detection of Protected Health Information (PHI) leaks (like SSN), and end-to-end sandbox integration testing. Ideal for deploying secure, compliant AI applications in the healthcare sector.

This workflow implements Abridge's end-to-end clinical documentation pipeline. It captures ambient audio from patient encounters, performs real-time transcription, utilizes generative AI to structure the medical notes (e.g., SOAP, H&P), generates required medical codes (ICD-10, CPT), and finally inserts the structured data directly into the EHR system. Designed for primary clinical encounters, minimizing manual charting time.

This comprehensive workflow guides the implementation of a complete Retrieval-Augmented Generation (RAG) pipeline using Cohere's core services. It demonstrates the full process: embedding documents, searching for relevant context, reranking results for precision, and finally generating a grounded answer with precise citations using the Chat API. Ideal for building knowledge-intensive, enterprise-grade Q&A systems.