Search local NVIDIA CUDA documentation for PTX and API references.
npx skills add https://github.com/slowlyC/agent-gpu-skills --skill cuda-skill-slowlyc
Please help me install this Agent Skill. Skill: cuda-skill Source: https://github.com/slowlyC/agent-gpu-skills/tree/main/cuda_skill Command: npx skills add https://github.com/slowlyC/agent-gpu-skills --skill cuda-skill-slowlyc
This Skill eliminates the need to manually search through scattered NVIDIA PDFs and online documentation for CUDA development. It provides instant, full-text search across locally stored PTX ISA 9.1, CUDA Runtime/Driver API 13.1, Programming Guide v13.1, Best Practices Guide, Nsight Compute, and Nsight Systems documentation directly from your AI agent.
Use the cuda-skill skill to search the local documentation for the exact syntax and usage constraints of the wgmma.mma_async PTX instruction on sm_90 GPUs.
High-intent search queries and answers about installing and using this skill.
Offline CUDA documentation search queries locally stored PTX ISA 9.1 and CUDA API reference files to retrieve exact instruction syntax and usage constraints without external internet connectivity or manual PDF navigation.
Looking up CUDA Runtime API error codes and device properties uses full-text search across local CUDA Runtime and Driver API 13.1 markdown files to provide deterministic, instant access to authoritative reference material.
Finding profiling workflows for Nsight Compute and Nsight Systems involves searching structured local documentation covering nsys and ncu tools to debug memory errors and optimize kernel performance.
Yes, CUDA documentation search covers architecture-specific features for Ampere, Hopper, and Blackwell GPUs, including WGMMA instructions and TMA swizzling modes for sm_90 GPUs found in the local PTX ISA documentation.
Yes, you can search the CUDA Best Practices Guide for memory management patterns because the locally stored documentation includes hundreds of markdown files covering memory management, debugging workflows, and kernel optimization techniques.
Using local CUDA documentation search eliminates manual PDF navigation by providing instant full-text search across PTX ISA, CUDA APIs, and profiling tool docs to ensure deterministic offline access to authoritative reference material.
carbonscottSearch local NVIDIA CUDA documentation for APIs and optimization guidance.
linkedinProfile, generate variants, and validate Liger Kernel Triton kernels for NVIDIA GPUs.
ariusewyDebug CUDA crashes in FlashInfer applications using API logging and compute-sanitizer.
dotruruLocate CUDA/PTX documentation and APIs in offline references.
