Install a full repository package, then let your agent activate the right skill for the task.
Ask your agent
Prompt
Fetch this Markdown manifest as plain text: https://skillscout.sh/official/nvidia/nvidia_skills_skillscout.md. If your browser cannot display the Markdown directly, use curl or another text fetcher. Recommend the relevant NVIDIA skills package or packages for my goal, explain which individual skills matter, and wait for my approval. After I approve, install with the npx command. If no project is open, install it as an agent skill I can use later.
Covers accelerated-computing-cudf, aiq-research, rag-blueprint, cuopt-developer, aiq-deploy, cudaq-guide. Diagnose and optimize OpenUSD scene performance Validate and use packed sequences and long-context training in Megatron-Bridge, distinguishing offline packed SFT for LLMs from in-batch packing for VLMs, and applying the right CP constraints.
Includes 11 official skills: nemo-relay-debug-runtime-integration, nemo-relay-instrument-calls, nemo-relay-migrate-from-flow, nemo-relay-get-started, nemo-relay-install, nemo-relay-instrument-context-isolation.
Includes 25 official skills: simready-foundation-create-package, simready-foundation-add-feature, simready-foundation-conform-fet-000-core, simready-foundation-conform-fet-003-rigid-body-physics, simready-foundation-conform-fet-006-materials, simready-foundation-conform-fet-023-robot-materials.
Includes 8 official skills: documentation-style, nvcf-explore-stack, nvcf-self-managed-cli, nvcf-self-managed-installation, nvcf-self-managed-prerequisite, nvca-chart-release.
Includes 12 official skills: nvflare-convert-pytorch, nvflare-diagnose-job, nvflare-orient, autofl-nvflare, autofl-nvflare-report, nvflare-convert-huggingface.
Validate and use packed sequences and long-context training in Megatron-Bridge, distinguishing offline packed SFT for LLMs from in-batch packing for VLMs, and applying the right CP constraints.
Covers accelerated-computing-cudf, aiq-research, rag-blueprint, cuopt-developer, aiq-deploy, cudaq-guide. Diagnose and optimize OpenUSD scene performance Validate and use packed sequences and long-context training in Megatron-Bridge, distinguishing offline packed SFT for LLMs from in-batch packing for VLMs, and applying the right CP constraints.
Top skills in this package
accelerated-computing-cudfOfficial NVIDIA skill from nvidia/skills.
aiq-researchOfficial NVIDIA skill from nvidia/skills.
rag-blueprintOfficial NVIDIA skill from nvidia/skills.
cuopt-developerOfficial NVIDIA skill from nvidia/skills.
aiq-deployOfficial NVIDIA skill from nvidia/skills.
cudaq-guideOfficial NVIDIA skill from nvidia/skills.
cuopt-installOfficial NVIDIA skill from nvidia/skills.
data-designerOfficial NVIDIA skill from nvidia/skills.