Skill UI

This skill guides developers on building LLM-powered applications using the Claude API, Anthropic SDKs, or the Agent SDK. It provides comprehensive guidance on language detection, best practices (like default model versions and streaming), and choosing the correct development surface (API call, Workflow, or Agent) based on the use case.

A comprehensive security tool designed to detect prompt injection attacks targeting Large Language Models (LLMs). It employs a multi-layered approach combining regex signature matching, structural anomaly scoring, and advanced DeBERTa transformer classification. It is essential for securing input validation layers in chatbots, AI agents, and RAG pipelines, and for retrospectively auditing LLM interaction logs against OWASP standards.

This skill detects AI-generated deepfake audio used in voice phishing (vishing) attacks. It functions by extracting critical spectral features (including MFCCs, spectral centroid, and zero-crossing rate) and passing them through advanced machine learning models (like Random Forest and Gradient Boosting). The system supports batch forensic analysis, provides confidence scores, and generates detailed reports to verify the authenticity of recorded voice calls, making it essential for cybersecurity incident response and fraud investigation.

A workflow that bridges experiment plans to execution. Reads an experiment plan, implements training and evaluation code, performs cross-model code review, deploys runs to GPU clusters, collects initial metrics, and prepares outputs for automated review loops.

This tool systematically structures and derives complex research formulas and theoretical models. It takes scattered equations, messy notes, and ambiguous ideas, transforming them into a coherent, paper-ready derivation package. Use it when you need to build a rigorous theory line, organize explicit assumptions, or clarify the conceptual object underlying a set of formulas.

This skill is designed for running comprehensive evaluations of Hugging Face Hub models on local hardware. It supports multiple evaluation frameworks, including `inspect-ai` and `lighteval`, and provides flexibility in backend selection (e.g., `vllm`, Transformers, or `accelerate`) to optimize for performance and compatibility. Use this for local GPU benchmarking, smoke testing, and backend selection, but note that it does not handle remote job orchestration or result publishing.

Gradio is a powerful Python library designed for creating interactive web user interfaces and showcasing machine learning demos quickly. This skill covers core patterns like Interface, Blocks, and ChatInterface, allowing users to build complex applications, components, and event-driven chatbots with minimal effort. Ideal for rapid prototyping and deep learning model visualization.

This skill provides comprehensive analysis and structured data extraction for academic research papers hosted on Hugging Face and arXiv. It can read paper content in markdown format, retrieve key metadata (authors, abstracts, GitHub repos), and identify linked assets such as associated models, datasets, and Spaces. Use this when the user provides a Hugging Face or arXiv URL, or explicitly requests a summary or explanation of an AI research topic.

This skill provides a managed cloud environment for training and fine-tuning advanced vision models. Users can build and customize models for object detection (e.g., DETR, YOLOS), image classification, and segmentation (SAM/SAM2). By leveraging Hugging Face Jobs, training is executed on powerful cloud GPUs without requiring local GPU setup, and all resulting models are automatically saved to the Hugging Face Hub.

This framework implements robust guardrails for LLM applications, validating both user inputs and model outputs. It detects prompt injection, sensitive PII (Personally Identifiable Information), toxic content, and ensures structured output compliance using advanced tools like NVIDIA NeMo Guardrails and Guardrails AI. Essential for building secure, reliable chatbots and RAG systems.

This skill provides continuous, structured monitoring of deep learning model training metrics. It analyzes data from platforms like WandB and local logs to detect critical anomalies such as NaN values, divergence, plateaus, and sudden spikes. By providing a clear decision (CONTINUE, WAIT, or STOP) and detailed evidence, it ensures compute resources are only spent on viable training runs.

Transformers.js allows developers to run state-of-the-art machine learning models directly within JavaScript/TypeScript environments (browser, Node.js). It supports a comprehensive range of tasks including NLP (text generation, summarization, translation), Computer Vision (image classification, object detection), and Audio processing. This capability enables building fully client-side, multimodal AI applications without requiring a dedicated Python backend or server infrastructure.