Skill UI

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 framework implements robust input and output validation guardrails for LLM-powered applications. It prevents prompt injection, detects and redacts PII, filters toxic content, and validates model outputs against defined schemas. Use it when deploying AI agents that handle sensitive data or require strict adherence to corporate policies, ensuring a secure and trustworthy AI interaction layer.

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.

Assesses AI/ML systems for prompt injection, jailbreak, model inversion, data poisoning, and agent tool abuse using MITRE ATLAS mapping, signature detection, and scoring metrics before deployment.

This skill guides users on setting up and running LLM-as-judge evaluations using Arize. It covers defining evaluators (including prompts, classification choices, and models) and executing tasks against real data spans or experiments. It supports key metrics like hallucination, faithfulness, and correctness, and facilitates continuous monitoring for comprehensive LLM performance analysis.

This skill provides a comprehensive framework for optimizing, improving, and debugging Large Language Model (LLM) prompts. It leverages production trace data, system evaluations, and human annotations to extract prompts from spans. Users can run a data-driven optimization loop, ensuring the AI responses meet desired quality standards. Ideal for prompt tuning, system prompt refinement, and improving overall AI output quality.

A comprehensive framework designed to structure and automate PyTorch deep learning workflows. It simplifies boilerplate code by organizing models into LightningModules, automating training loops via the Trainer, and providing robust support for multi-GPU/TPU scaling (DDP, FSDP, DeepSpeed), data handling (DataModules), and best-in-class experiment logging (W&B, MLflow). Ideal for professional, scalable deep learning project development.

A comprehensive Python framework utilizing deep generative models and variational inference for advanced single-cell genomics analysis. It supports the integration and analysis of diverse multi-modal data types, including single-cell RNA-seq, ATAC-seq, CITE-seq, and spatial transcriptomics. Key applications include probabilistic batch correction, cell type annotation, and multi-omics integration.

This guide covers advanced usage of the Anthropic Messages API. Learn how to implement real-time streaming responses, utilize system prompts for persona control, process visual inputs (Vision), and enforce structured data output (JSON). This is essential for building sophisticated, robust, and multi-modal Claude-powered applications.

A comprehensive security auditing framework designed to systematically detect sensitive information disclosure from Large Language Models (LLMs). It uses multi-phase, escalating dialogue probes to test for leaks such as system prompts, API keys, PII, and internal configurations. Essential for red teaming and rigorous model security assessment.

This comprehensive skill provides multi-layered guardrails for integrating with LLMs like Claude. It enforces security and compliance by implementing input validation (PII detection, length checks), system prompt restriction (scope control, safety rules), output filtering (preventing sensitive data leakage), cost governance, and model access policies. Ideal for building secure, robust, and governed AI applications.

Learn how to build complex, multi-step AI agents using Cohere's Chat API v2. This skill demonstrates defining external tools (like weather APIs or database searches) and implementing the full agent loop: model calls tool -> execute tool -> feed result back -> final answer. Ideal for creating autonomous, tool-using workflows that interact with external systems.