LLM Engineering Bundle

Skills for building, training, and deploying large language models.

Browse all skill bundles

Included Skills (100)

  1. Nemo Automodel Recipe Development — Create and modify NeMo AutoModel training and evaluation recipes, including YAML structure, builders, and execution flow.
  2. Earth2studio Data Fetch — Fetch weather/climate data via Earth2Studio data sources for specific variables and times. Do NOT use for inference pipelines, model discovery, or installation.
  3. CLIP Model Skill — A skill for fine-tuning, classifying, and deploying CLIP vision-language models.
  4. Nemo Automodel Distributed Training — Guide for selecting and configuring distributed training strategies in NeMo AutoModel, including FSDP2, Megatron FSDP, DDP, and parallelism settings.
  5. Kermt Finetune — Finetune a pretrained KERMT encoder on a labeled CSV. Validate the checkpoint and data, prepare features, and run containerized training. Use a local checkpoint or optionally download a pinned Hugging Face model bundle using HF_TOKEN if configured. Write model bundles, prepared data, logs, and trained models to user-selected host directories.
  6. M365 Agent SDK — This Python SDK enables developers to build and host enterprise-grade multichannel agents for Microsoft Teams, Microsoft 365, and Copilot Studio environments.
  7. Agent Development — This skill should be used when the user asks to "create an agent", "add an agent", "write a subagent", "agent frontmatter", "when to use description", "agent examples", "agent tools", "agent colors", "autonomous agent", or needs guidance on agent structure, system prompts, triggering conditions, or agent development best practices for Claude Code plugins.
  8. Agent Platform Prompt Management — Manages and orchestrates prompts in Agent Platform. Use when you need to create, list, retrieve, version, or delete managed prompts in Agent Platform. Don't use for model training, model deployment to endpoints, or managing non-Agent Platform prompts.
  9. Agent Platform Alert Configuration — Configures best-practice alerting policies for AI agents using OpenTelemetry (OTel) metrics, generating output as Terraform (.tf) configuration files. Use when analyzing, writing, or deploying alerting policies to monitor agent latency, error rates, token usage, and quality metrics. Don't use for standard infrastructure monitoring unrelated to AI agents, or when the agent is not instrumented with OpenTelemetry (for Reliability, Cost, Safety, Security alerts). NOTE: Reliability, Cost, Safety, and Security alerts use generic OTel metrics and work across runtimes (such as Cloud Run, Vertex AI). Quality alerts rely on Vertex AI Online Monitors and are strictly bound to Vertex AI deployments.
  10. Cdf Curation And Retrieval — NVIDIA CDF curation and retrieval. Cosmos Curator video and image pipelines (split, dedup, shard, annotate) with filtering, captioning, and embedding. Cosmos Dataset Search ingest, search, and capability qualification. TAO data mining embeddings and neighbor matching. Ships domain event catalogs, including workplace employee conduct monitoring, which requires legal review before use. Use when a request names Cosmos Curator, Cosmos Dataset Search, CDF, or their Make and CLI surfaces. Not for model training, orchestration, or unrelated dataset questions.
  11. Tao Run Deft Aoi Cosmos3 — Run the disk-backed DEFT AOI improvement loop for NVIDIA Cosmos Reason 3 / Cosmos3 models, using Nano by default and Edge or Super when explicitly requested: evaluate the base model on Proxy and frozen Benchmark splits, mine real image pairs from Proxy gaps, generate AnomalyGen synthetic NG pairs, assemble a per-iteration Train JSON from both producers, train with Cosmos Framework LoRA SFT, and repeat through the selected platform's submit/status/logs/cancel contract. This migration supports bare labels only: the assistant response must be exactly OK or NG. Use for "run Cosmos3 DEFT AOI", "CR3 AOI loop", or "improve Cosmos3 PCB inspection with bare OK/NG"; do not use for rich/reasoning annotation, one-off Cosmos training, or generic anomaly generation.
  12. Kermt Continue Pretrain — Continue KERMT pretraining on a custom SMILES corpus with a grover_base, cmim, or hybrid checkpoint. Use a local checkpoint or optionally download a pinned Hugging Face model bundle using HF_TOKEN if configured. Run containerized training and write model bundles, prepared data, logs, and checkpoints to user-selected host directories.
  13. Nemo Mbridge Perf Sequence Packing — Validate and use packed sequences and long-context training in Megatron-Bridge, including offline LLM packing, collate-time VLM packing, Energon online packing, and CP constraints.
  14. Codonfm Setup — Set up the public CodonFM v1 repository and download public Encodon checkpoints. Use for requests to build or launch the CodonFM development container, configure local data/checkpoint mounts, verify GPU access, or download public Encodon 80M, 600M, 1B, or Cdwt-1B weights. Do not use for Decodon, Encodon 5B/10B, missense-aggregation, or codon-optimization setup because those implementations are not in the public repository.
  15. Nemo Mbridge Perf Expert Parallel Overlap — Validate and use MoE expert-parallel communication overlap in Megatron-Bridge, including overlap_moe_expert_parallel_comm, delay_wgrad_compute, and flex dispatcher backends such as DeepEP and HybridEP.
  16. Nemo Automodel Model Onboarding — Guide for onboarding new model architectures into NeMo AutoModel, including architecture discovery, implementation patterns, registration, and validation.
  17. Rtvi Cv Customize Model — How to swap the DeepStream CV detection model in the VSS Alerts Blueprint verification (2d_cv) mode - covers ONNX export, custom bbox parsers, compose mount gotchas, nvinfer config, runtime TRT engine build, deployment, and a segmentation-capable model addendum handoff.
  18. Video CLIP Trainer — Fine-tunes InternVideo2-CLIP for video-text retrieval and classification, helping developers build multimodal AI models.
  19. Tao Train Codetr — Co-DETR (CoDINO) for object detection. A DETR-family detector with collaborative hybrid
  20. Mcore Testing — Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.
  21. Ethos Write — Ethos Write creates a comprehensive ethos.md file for an AI agent, capturing its purpose, behavior guidelines, boundaries, and success criteria based on data from ethos‑explore. It validates the schema, ensures all required sections are present, saves the file to the agent’s repository, and presents it for user review. Existing ethos.md files can also be edited, providing a reliable, reusable framework for agent governance.
  22. LLM Prompt Optimizer — This tool refines LLM prompts using production trace data and performance metrics to help developers improve AI output quality through data-driven optimization.
  23. ContextMatic Onboarding Guide — This interactive tour helps new users understand the ContextMatic MCP server by exploring available APIs and demonstrating core search functionalities through guided conversation.
  24. Arize AI Integration — This skill manages LLM provider credentials within Arize, enabling developers to securely configure and maintain integrations for evaluators and downstream AI features.
  25. Nemo Mbridge Perf Memory Tuning — Techniques for reducing peak GPU memory in Megatron Bridge — expandable segments, PEFT + SP input re-gather, parallelism resizing, activation recompute, CPU offloading constraints, and common OOM fixes.
  26. Azure OpenAI Client — This .NET client library enables developers to integrate Azure OpenAI and OpenAI services for chat, embeddings, image generation, and audio transcription tasks.
  27. LLM Context Updater — This tool updates the repository's llms.txt file to ensure AI models have accurate, compliant, and up-to-date guidance for navigating project documentation.
  28. Tao Run Automl — Run container-backed AutoML / hyperparameter optimization (HPO) for NVIDIA TAO networks using AutoMLRunner. Handles algorithm
  29. LLMs.txt Generator — Generates a compliant llms.txt file from repository structure to help large language models effectively navigate and understand project documentation and codebase specifications.
  30. Context Engineering — Optimizes agent context setup. Use when starting a new session, when agent output quality degrades, when switching between tasks, or when you need to configure rules files and context for a project.
  31. Draw.io Diagram Generator — This tool generates and edits structured Draw.io files, enabling developers and architects to visualize complex systems, workflows, and data models with ease.
  32. Nemo Mbridge Perf Cuda Graphs — Validate and use CUDA graph capture in Megatron Bridge, including local full-iteration graphs and Transformer Engine scoped graphs for attention, MLP, and MoE modules.
  33. M365 Agents Dotnet — The M365 Agents Dotnet skill supplies a .NET SDK that lets developers build multichannel agents for Microsoft Teams, M365, and Copilot Studio. It includes ASP.NET Core hosting, AgentApplication routing, and MSAL-based authentication, enabling secure and scalable agent applications. Software engineers and enterprise developers use it to embed intelligent agents directly into Microsoft’s collaboration ecosystem.
  34. Deepstream Import Vision Model — Use this skill to bring a supported object-detection vision model from HuggingFace or NVIDIA NGC into an NVIDIA DeepStream pipeline with end-to-end automation: ONNX download, SafeTensors export, TRT engine build, custom nvinfer bbox parser, multi-stream benchmark, and PDF report. Object detection models only.
  35. Autonomous Experimentation Loop — This tool automates iterative programming tasks by guiding developers through goal setting, code modification, and performance measurement to optimize software through continuous experimentation.
  36. Warp Debug Gradients — Use to diagnose and fix incorrect gradients in differentiable Warp programs. Anything trained, optimized, calibrated, or fit through Warp kernels depends on wp.Tape gradients, so treat any misbehavior of such a workflow as a gradient problem until proven otherwise — use this when training diverges or NaNs, won't train at all, stalls or plateaus above the expected loss, converges to a wrong or biased answer, is worse than a reference implementation, works at small scale but fails at production scale, or fails a QA/validation recheck. Also for explicit symptoms — exploding, NaN/inf, zero, or subtly wrong gradients, suspected wp.Tape/backward issues, gradcheck failures — but users usually describe only the surface symptom ("the sim explodes", "the fit gets dragged toward outliers") without mentioning gradients: make that leap. Not for forward-only Warp work, build/install problems, or autograd issues in other frameworks without Warp.
  37. Nvflare Convert Huggingface — Convert existing Hugging Face Transformers Trainer or TRL SFTTrainer training code into an NVFLARE federated job using flare.patch(trainer), local validation, and job export; use when the user names Hugging Face or preliminary source inspection identifies one Hugging Face owner, and not for manual PyTorch loops, Lightning, inference-only pipelines, deployment, or experiment workflows.
  38. Tao Finetune Cosmos Reason — Shared Cosmos3 frontend that explicitly routes Cosmos Framework and Cosmos-RL, validates runtime model/video-dataset/SLURM inputs, consumes an SQSH or packaged backend image, optionally plans explicit clean source builds, prepares checkpoints, validates the first update in-process, and returns token-weighted losses and task-aware accuracy.
  39. Google Cloud Solution Agentic AI Data Science Workflow — Designs a tailored multi-product agentic data science architecture on Google Cloud that incorporates opinionated best practices. Use when architecting multi-product solutions for agent-based data analytics or ML workloads. Don't use for simple queries, non-agentic pipelines, general cloud reviews, or writing agent code.
  40. Earth2studio Install — Guide installing Earth2Studio via uv or pip, selecting model extras, and configuring the environment. Do NOT use for writing inference code, choosing models, or PhysicsNeMo questions.
  41. Java Idiomatic Reference — This guide provides modern, idiomatic Java coding standards and best practices to help developers write cleaner, more efficient, and maintainable application code.
  42. Cirq — Cirq is Google Quantum AI's open-source framework for designing, simulating, and running quantum circuits on quantum computers and simulators.
  43. llms.txt Generator — Creates and analyzes llms.txt files to help AI systems understand website structure and content.
  44. M365 Agents Dotnet — Microsoft 365 Agents SDK for .NET. Build multichannel agents for Teams/M365/Copilot Studio with ASP.NET Core hosting, AgentApplication routing, and MSAL-based auth.
  45. Performing Wifi Password Cracking With Aircrack — Captures WPA/WPA2 handshakes and performs offline password cracking
  46. Phishing Simulation Orchestrator — This tool executes authorized phishing campaigns to help security professionals measure organizational susceptibility to social engineering and improve overall human security awareness.
  47. Prompt Template Library — This collection provides developers and users with high-quality, role-based prompt templates to optimize AI interactions and improve task-specific output accuracy.
  48. Bash Terminal Assistant — This tool provides essential Bash and Linux command patterns to help developers efficiently manage files, processes, and system operations in terminal environments.
  49. Go Idiomatic Reference — This guide provides Go developers with essential best practices and idiomatic patterns to ensure efficient, clean, and maintainable code across projects.
  50. Nodejs Best Practices — Node.js development principles and decision-making. Framework selection, async patterns, security, and architecture. Teaches thinking, not copying.
  51. Env Secrets Manager — Manage environment-variable hygiene and secrets safety across local development and production. Practical auditing, drift awareness, rotation readiness. Use when auditing .env files for committed secrets, planning a credential rotation, debugging missing-env-var production incidents, or hardening a new project against secrets leakage.
  52. Skill Router — Use when the user is unsure which skill to use or where to start. Interviews the user with targeted questions and recommends the best skill(s) from the installed library for their goal.
  53. Llm Evaluation — Master comprehensive evaluation strategies for LLM applications, from automated metrics to human evaluation and A/B testing.
  54. Loopy — Discover, find, compare, audit, repair, adapt, craft, run, debrief, and prepare repeatable AI-agent loops for publication.
  55. Referral Program — When the user wants to design, launch, or optimize a referral or affiliate program. Use when they mention 'referral program,' 'affiliate program,' 'word of mouth,' 'refer a friend,' 'incentive program,' 'customer referrals,' 'brand ambassador,' 'partner program,' 'referral link,' or 'growth through referrals.' Covers program mechanics, incentive design, and optimization — not just the idea of referrals but the actual system.
  56. Vertex Media Generation — Generates images with Gemini image models and videos with Veo through the Vercel AI SDK Google Vertex provider. Use when the user wants to generate or edit images, generate videos, migrate off discontinued Imagen models, or build media pipelines with @ai-sdk/google-vertex. Covers Gemini image models and Veo 3.1.
  57. Fitness Analyzer — 分析运动数据、识别运动模式、评估健身进展，并提供个性化训练建议。支持与慢性病数据的关联分析。
  58. Diagram Generator — Generate, refine, validate, and render diagrams from natural language, notes, code, schemas, or existing diagram sources: flowcharts, swimlanes, attack-path graphs, data-flow diagrams, architecture, and state machines.
  59. Real-Time Agent Orchestrator — This tool provides comprehensive documentation for building low-latency, multimodal AI applications using the Pipecat framework and Daily infrastructure for developers.
  60. AI Eval Ci — AI Eval Ci integrates AI agent and LLM evaluations directly into CI/CD pipelines, automatically failing builds when output quality falls below defined thresholds. It supports frameworks such as Cobalt, Promptfoo, and Braintrust, and offers LLM-as-judge scoring and threshold assertions, all wrapped in GitHub Actions for seamless deployment. This tool benefits developers and data scientists who need to catch prompt regressions, benchmark models, and enforce quality gates before releasing AI systems.
  61. Agent Tool Architect — Design and optimize effective tool APIs for AI agents to ensure reliable interaction, better performance, and reduced failure rates during execution.
  62. Codex CLI Bridge — Bridge between Claude Code and OpenAI Codex CLI - generates AGENTS.md from CLAUDE.md, provides Codex CLI execution helpers, and enables seamless interoperability between both tools
  63. Hugging Face Papers — This tool retrieves and parses Hugging Face research paper metadata, helping AI developers quickly access linked models, datasets, and project repositories.
  64. Performing Soc Tabletop Exercise — Performs tabletop exercises for SOC teams simulating security incidents
  65. Ragas — Ragas is an open-source Python framework for evaluating RAG pipelines and LLM applications with metrics such as faithfulness, response relevancy, context precision and context recall. Use when asked to evaluate or score a RAG system, generate a synthetic test set from documents, add RAG quality checks to CI, or write a custom LLM-judged metric.
  66. Stitch Prompt Engineer — This skill provides expert guidance for crafting precise, actionable prompts to help designers and developers generate high-quality UI layouts using Google Stitch.
  67. Efficient Code Architect — Enforces a concise, idiomatic, and maintainable coding standard to minimize token bloat and operational overhead for developers during all software generation tasks.
  68. Tao Train Mask Auto Encoder — Masked Auto-Encoder (MAE) for self-supervised pretraining and fine-tuning. Masks random patches and reconstructs
  69. Medical Evidence Exporter — This tool exports sanitized medical AI inference metadata and quality metrics to MLflow, assisting researchers in maintaining reproducible model evidence records.
  70. Vibe Coding — Patterns and workflows for AI-assisted development (vibe coding) — building software primarily through natural language with AI coding agents. Use when: setting up an AI-first development workflow, structuring projects for AI coding, using prompt-driven development best practices, or maximizing output from Claude Code, Cursor, Windsurf, or Aider.
  71. Azure OpenAI Client — This library enables .NET developers to integrate Azure OpenAI and OpenAI services for chat, embeddings, image generation, and advanced assistant workflows.
  72. Google Agents CLI Onboarding — Onboarding entrypoint for agents-cli in Agent Platform. It should be used when the user wants to "create a new agent", "develop an agent", "build an agent using ADK", "run the agent locally", "debug agent code", "test an agent", "evaluate an agent", "deploy an agent", "publish an agent", "monitor an agent", or needs the ADK (Agent Development Kit) development lifecycle.
  73. Developer Audience Context — When the user wants to establish or update their developer audience context. Also use when starting any other developer marketing skill to ensure foundational context is loaded.
  74. Prompt Tester — The Prompt Tester designs, evaluates, and refines AI prompts using structured criteria to build features and optimize agent instructions. It compares variants to assess performance across edge cases, ensuring reliable prompts. Developers and data scientists use it to enhance AI system performance.
  75. Scope Guardrail Assistant — This tool enforces strict project scope and prevents over-engineering by ensuring AI agents only modify files explicitly requested by the user.
  76. System Prompt Auditor — Identifies sensitive information leakage in LLM system prompts for security professionals conducting red-team assessments and validating application authorization logic.
  77. Python Idiomatic Guide — This skill provides developers with essential best practices and idiomatic patterns to write cleaner, more efficient, and highly maintainable Python code.
  78. Agent Creator — Create custom AI subagents with proper plugin structure, persona generation, and companion routing skills.
  79. Mcore Run On Slurm — How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.
  80. Context Management Framework — This framework helps developers maintain structured project context as a first-class artifact to ensure consistent AI interactions and improved team alignment during development.
  81. Cost Optimizer — Helps developers reduce LLM API and infrastructure costs through strategic optimization.
  82. Codonfm Embed — Validate coding-sequence CSVs, extract public CodonFM/Encodon embeddings, and choose checkpoints for downstream property modeling.
  83. Agent Orchestration Improve Agent — Systematic improvement of existing agents through performance analysis, prompt engineering, and continuous iteration.
  84. Claude Hud — Build heads-up display (HUD) dashboards for AI coding agents — show context usage, active tools, running sub-agents, and task progress in real-time. Use when: monitoring AI agent activity, building developer tools for AI-assisted coding, creating status dashboards for agent workflows.
  85. Hugging Face Jobs — Run workloads on Hugging Face Jobs with managed CPUs, GPUs, TPUs, secrets, and Hub persistence.
  86. Nvidia Skill Finder — Use for NVIDIA-related requests where an NVIDIA skill might help, even if the user did not ask for a skill. Trigger on NVIDIA products, hardware, software, SDKs, GPUs, Jetson/JetPack/L4T/BSP/SDK Manager/driver/flashing/setup, CUDA, NIM, NeMo, Omniverse/OpenUSD/SimReady, Isaac, RAPIDS/cuDF, cuPyNumeric, cuOpt, Dynamo, Holoscan, TensorRT, DeepStream, VSS, TAO, NGC/NVCF. Do not use for generic non-NVIDIA route, optimize, deploy, AI, video, data, or infrastructure tasks.
  87. C# Idiomatic Guide — This skill provides developers with best practices and idiomatic coding patterns to write efficient, clean, and maintainable C# source code.
  88. Faf Wizard — Done-for-you .faf generator. One-click AI context for any project - new, legacy, or famous. Auto-detects stack, scores readiness, works everywhere.
  89. Claude Skill Architect — This guide assists developers in building and managing Claude Code skills by implementing best practices like the 500-line rule and progressive disclosure patterns.
  90. Hugging Face Community Evals — Run evaluations for Hugging Face Hub models using inspect-ai and lighteval on local hardware. Use for backend selection, local GPU evals, and choosing between vLLM / Transformers / accelerate.
  91. Llm Caching — Implement multi-layer LLM caching with exact match, semantic similarity,
  92. Memory Systems — Design short-term, long-term, and graph-based memory architectures. Use when building agents that must persist across sessions, needing to maintain entity consistency across conversations, or implementing reasoning over accumulated knowledge.
  93. Earth2studio Discover — Find Earth2Studio models, data sources, and examples for a weather/climate use case. Do NOT use for writing inference code, downloading data, or installation.
  94. Jetson Package — Pick Jetson-compatible containers, vLLM runtime images, and Jetson AI Lab PyPI indexes; maps Orin SM 8.7 vs Thor SM 11.0 and JetPack-specific package choices.
  95. Huggingface Zerogpu — AI demos and GPU compute with Gradio Spaces and Hugging Face Spaces ZeroGPU.
  96. Makepad Shaders — CRITICAL: Use for Makepad shader system. Triggers on: makepad shader, makepad draw_bg, Sdf2d, makepad pixel, makepad glsl, makepad sdf, draw_quad, makepad gpu, makepad 着色器, makepad shader 语法, makepad 绘制
  97. Kaggle Finetune — End-to-end workflow for fine-tuning LLMs using Kaggle datasets. Use when downloading datasets from Kaggle for model training, preparing conversation/customer service data for chatbot fine-tuning, or building domain-specific AI assistants. Covers dataset discovery, download, preprocessing into chat format, and integration with PEFT/LoRA training.
  98. Realtime Voice Assistant — This skill helps developers build voice-enabled AI applications using the OpenAI Realtime API for real-time voice conversations.
  99. Loop Library — Discover, find, compare, audit, repair, adapt, and design repeatable AI-agent loops with explicit triggers, actions, verification, stopping conditions, guardrails, and handoffs. Use when a user asks to analyze a codebase for potential loops, mine coding-thread history for work done more than once, turn repeated engineering work into a loop, find or recommend a published loop, create a recurring agent workflow or automation cadence, turn an outcome into a bounded copy-ready loop, or review an existing loop for weak checks, unsafe authority, unbounded repetition, stale state, or unclear stopping behavior.
  100. Vercel Deployment Manager — Helps developers deploy and manage full-stack applications on Vercel with previews and guardrails.

Opening your workspace…