Skill UI

llama.cpp is a pure C/C++ library designed for efficient LLM inference, optimized to run on resource-constrained environments. It excels in scenarios where high-end NVIDIA GPUs (CUDA) are unavailable, making it ideal for Apple Silicon Macs, AMD/Intel GPUs, or edge deployment devices like Raspberry Pi. It supports GGUF quantization formats, enabling significant memory reduction and speed improvements for local, cross-platform AI applications.

NVIDIA’s NeMo Guardrails adds runtime safety rails to LLM apps, covering jailbreak detection, hallucination and fact-checking, PII filtering, and toxicity checks via Colang 2.0 rules and integrations for production use on T4 GPUs.

Optimizes large language model inference on NVIDIA GPUs with TensorRT, delivering 10-100× faster throughput, quantized precision (FP8/INT4), multi-GPU scaling, and serving-ready tooling for production deployments.

Collects comprehensive diagnostic evidence for troubleshooting Vast.ai GPU instances. This bundle gathers account status, lists all running instances, retrieves recent logs, performs remote GPU diagnostics (e.g., nvidia-smi, CUDA checks), and verifies API connectivity. It is essential for creating detailed support tickets when encountering persistent infrastructure issues.

Run NVIDIA Cosmos Policy evaluations on LIBERO and RoboCasa simulations, enabling headless EGL GPU rendering, inference profiling, and benchmark workflows via public eval entrypoints.

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 assistant provides comprehensive performance profiling for complex computing systems. It analyzes various targets, including Python scripts, running processes, GPU utilization, and memory bottlenecks. Utilizing both external tools (like nvidia-smi, cProfile) and advanced code instrumentation, it measures CPU, memory, and interconnect performance, generating a structured report with actionable insights and a mandatory changelog of all modifications.

Accelerates CPU-bound Python workflows by translating NumPy, pandas, ML, graph, image, simulation, and IO code into NVIDIA GPU kernels using CuPy, Numba CUDA, Warp, cuDF/cuML/cuGraph, KvikIO, cuSpatial, and RAFT for dramatic speedups.

Accelerate scientific Python code on NVIDIA hardware using CUDA and RAPIDS libraries. Optimizes NumPy, pandas, scikit-learn, and ML workloads for GPU execution. Includes profiling, memory management, and kernel optimization to ensure correctness and performance gains.

This skill provides comprehensive guidance for deploying and managing GPU-accelerated AI inference services on CoreWeave Kubernetes (CKS). It covers best practices including containerization using NVIDIA CUDA base images, configuring specific GPU resource limits (A100/H100), setting up robust health checks, and executing controlled rolling updates. Ideal for managing multi-model inference and scaling demanding AI workloads in a cloud environment.

NVIDIA garak is an open-source LLM vulnerability scanner designed for comprehensive red-teaming and security assessment. It systematically probes large language models (via OpenAI-compatible APIs, Hugging Face, etc.) using thousands of adversarial prompts. It covers critical vulnerabilities such as prompt injection, jailbreaking, data leakage, and toxicity. Use this tool to establish a repeatable security baseline before or after deployment, validate the effectiveness of safety guardrails, or generate quantitative evidence for AI risk assessments.

This skill provides comprehensive techniques for security-testing Retrieval-Augmented Generation (RAG) pipelines. It probes for indirect prompt injection, which occurs when malicious instructions are embedded in retrieved documents, and tests for embedding-space manipulation. Using advanced tools like NVIDIA garak, Promptfoo, and PyRIT against vector stores (FAISS, Chroma, etc.), users can validate retrieval guardrails and secure CI/CD gates against adversarial inputs and data leakage.