Claude Skills

LlamaIndex is a data framework for building RAG-powered LLM applications, specializing in document ingestion, indexing, and querying. It provides key features like vector indices, query engines, and agents, and supports over 300 data connectors. Use it for document Q&A, chatbots, and knowledge retrieval when building data-centric applications.

LangChain is a framework for building LLM applications using agents, chains, and RAG pipelines. It supports multiple LLM providers, offers 500+ integrations, and includes features like tool calling and memory management. Use it for rapid prototyping and deploying production systems like chatbots, autonomous agents, and question-answering services.

This Claude Skill serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. It's ideal for deploying production LLM APIs, optimizing inference performance, or serving models with limited GPU memory. The skill supports OpenAI-compatible endpoints, multiple quantization methods, and tensor parallelism.

TensorRT-LLM is an NVIDIA library that optimizes LLM inference for maximum throughput and lowest latency on NVIDIA GPUs. It is ideal for production deployments requiring 10-100x faster performance than PyTorch, supporting features like quantization and multi-GPU scaling. Use it when you need top performance on NVIDIA hardware, opting for alternatives like vLLM for simpler setups or llama.cpp for CPU/Apple Silicon.

SGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.

llama-cpp enables efficient LLM inference on non-NVIDIA hardware including CPUs, Apple Silicon, and consumer GPUs. It's ideal for edge deployment, Macs, or when CUDA is unavailable, offering GGUF quantization for reduced memory usage. This provides 4-10× speedup over PyTorch on CPU with minimal dependencies.

This Claude Skill runs the lm-evaluation-harness to benchmark LLMs across 60+ standardized academic tasks like MMLU and GSM8K. It's designed for developers to compare model quality, track training progress, or report academic results. The tool supports various backends including HuggingFace and vLLM models.

This Claude Skill provides expert guidance for PyTorch Fully Sharded Data Parallel (FSDP) training, helping developers implement distributed training solutions. It covers key features like parameter sharding, mixed precision, CPU offloading, and FSDP2 for large-scale model training. Use this skill when working with FSDP APIs, debugging distributed training code, or learning best practices for sharded data parallelism.

This Claude Skill trains large language models (2B-462B parameters) using NVIDIA's Megatron-Core framework with advanced parallelism strategies. Use it when training models over 1B parameters, needing maximum GPU efficiency (47% MFU on H100), or requiring tensor/pipeline/sequence parallelism. It's a production-ready framework proven on models like Nemotron, LLaMA, and DeepSeek.

This skill provides expert guidance for distributed training using Microsoft's DeepSpeed library. It helps developers implement optimization techniques like ZeRO stages, pipeline parallelism, and mixed-precision training. Use this skill when working with DeepSpeed features, debugging code, or learning best practices for large-scale model training.

HuggingFace Accelerate provides the simplest API for adding distributed training to PyTorch scripts with just 4 lines of code. It offers a unified interface for multiple distributed training frameworks like DeepSpeed, FSDP, and DDP while handling automatic device placement and mixed precision. This makes it ideal for developers who want to quickly scale their PyTorch training across multiple GPUs or nodes without complex configuration.

nemo-guardrails is NVIDIA's runtime safety framework for LLM applications that adds programmable safety rails. It provides jailbreak detection, input/output validation, fact-checking, toxicity detection, and PII filtering using Colang 2.0 DSL. Use this skill when you need production-ready safety controls for your LLM applications running on standard GPU hardware.

LlamaGuard is Meta's 7-8B parameter model for moderating LLM inputs and outputs across six safety categories like violence and hate speech. It offers 94-95% accuracy and can be deployed using vLLM, Hugging Face, or Amazon SageMaker. Use this skill to easily integrate content filtering and safety guardrails into your AI applications.

This skill provides expert guidance for fine-tuning LLMs using LLaMA-Factory, a framework featuring a no-code WebUI and support for 100+ models. It offers comprehensive assistance for implementing solutions, debugging code, and learning best practices when working with LLaMA-Factory's capabilities like multi-bit QLoRA and multimodal support. Use this skill when developing, debugging, or asking about LLaMA-Factory features and APIs.

Statistical analysis toolkit. Hypothesis tests (t-test, ANOVA, chi-square), regression, correlation, Bayesian stats, power analysis, assumption checks, APA reporting, for academic research.

Write scientific manuscripts. IMRAD structure, citations (APA/AMA/Vancouver), figures/tables, reporting guidelines (CONSORT/STROBE/PRISMA), abstracts, for research papers and journal submissions.

Evaluate research rigor. Assess methodology, experimental design, statistical validity, biases, confounding, evidence quality (GRADE, Cochrane ROB), for critical analysis of scientific claims.

Research ideation partner. Generate hypotheses, explore interdisciplinary connections, challenge assumptions, develop methodologies, identify research gaps, for creative scientific problem-solving.

Systematic framework for evaluating scholarly and research work based on the ScholarEval methodology. This skill should be used when assessing research papers, evaluating literature reviews, scoring research methodologies, analyzing scientific writing quality, or applying structured evaluation criteria to academic work. Provides comprehensive assessment across multiple dimensions including problem formulation, literature review, methodology, data collection, analysis, results interpretation, and scholarly writing quality.

Systematic peer review toolkit. Evaluate methodology, statistics, design, reproducibility, ethics, figure integrity, reporting standards, for manuscript and grant review across disciplines.

Conduct comprehensive, systematic literature reviews using multiple academic databases (PubMed, arXiv, bioRxiv, Semantic Scholar, etc.). This skill should be used when conducting systematic literature reviews, meta-analyses, research synthesis, or comprehensive literature searches across biomedical, scientific, and technical domains. Creates professionally formatted markdown documents and PDFs with verified citations in multiple citation styles (APA, Nature, Vancouver, etc.).

Spreadsheet toolkit (.xlsx/.csv). Create/edit with formulas/formatting, analyze data, visualization, recalculate formulas, for spreadsheet processing and analysis.

Presentation toolkit (.pptx). Create/edit slides, layouts, content, speaker notes, comments, for programmatic presentation creation and modification.

PDF manipulation toolkit. Extract text/tables, create PDFs, merge/split, fill forms, for programmatic document processing and analysis.