Trusted skills for AI agents from official vendors and established maintainers. Every skill is attributed to its author, scanned, and comes with install steps for your AI tool.
155 skills
NVIDIAOfficial
A guide to diagnosing and resolving GPU out-of-memory errors in Megatron Bridge training jobs.
NVIDIAOfficial
A troubleshooting guide for resolving runtime integration issues with NeMo Relay.
NVIDIAOfficial
Use when adding, modifying, optimizing, or debugging CuTile autotuning code. Trigger signals: `exhaustive_search` / `replace_hints` / `hints_fn` / `cuda.tile.tune` in code, `autotune` in filenames, or correctness/performance issues in autotuned CuTile kernels. Covers: tune-once/cache/launch pattern, per-architecture configs (sm80–sm120), parameter space design (tile sizes, occupancy, num_ctas), and 7 common pitfalls with solutions.
NVIDIAOfficial
A systematic workflow for diagnosing and resolving failures in Holoscan CLI commands.
NVIDIAOfficial
Technical guidance for configuring and optimizing Mixture-of-Experts (MoE) models for long-context training.
NVIDIAOfficial
Guide for using the pcc_counters.sh script to read firmware-level PCC diagnostic counters on NVIDIA ConnectX and BlueField devices.
NVIDIAOfficial
Diagnose NVSHMEM runtime failures and generate structured bug reports.
NVIDIAOfficial
Provides guidance for using doca_flow_tune to analyze, visualize, and optimize DOCA Flow pipelines.
NVIDIAOfficial
Profile a DeepStream pipeline with Nsight Systems and derive its configs from the measurement. Use when the user asks for an efficient, performant, or profiled pipeline — or to benchmark, tune, or measure FPS.
NVIDIAOfficial
Provides guidance for running the doca_perftest RDMA benchmark tool to measure network performance metrics.
NVIDIAOfficial
Evaluate if a code bottleneck is a suitable candidate for NVIDIA Warp acceleration.
NVIDIAOfficial
Configure and validate expert-parallel communication overlap for MoE models in Megatron-Bridge.
NVIDIAOfficial
Detect and diagnose CUDA correctness bugs with NVIDIA Compute Sanitizer. Routes GPU memory errors, shared-memory data races, uninitialized-memory reads, and synchronization/barrier bugs to the right sub-tool (memcheck, racecheck, initcheck, synccheck) and explains how to build, run, and read the report. Use when debugging illegal memory access, out-of-bounds, misaligned access, leaks, WAW/WAR/RAW races, __shared__ hazards, uninitialized reads, garbage results, hangs, divergent __syncthreads/__syncwarp, or cuda::barrier / wgmma issues. Do NOT use for host/CPU-side bugs (use valgrind or AddressSanitizer) or global-memory data races (out of scope). Trigger keywords - compute-sanitizer, memcheck, racecheck, initcheck, synccheck, CUDA memory error, OOB, race condition, uninitialized memory, __syncthreads, barrier, deadlock.
NVIDIAOfficial
Diagnose terminal failures, crashes, and startup errors for Run:ai workloads.
NVIDIAOfficial
Guidelines for maintaining isolated scope stacks in concurrent NeMo Relay environments.
NVIDIAOfficial
Analyze and debug network performance on NVIDIA hardware using the DOCA ngauge tool.
NVIDIAOfficial
Diagnose and repair specific NVIDIA PVA operator failures using evidence-based debugging workflows.
NVIDIAOfficial
Inventory host-DPU DOCA Comm Channel servers and connections using the doca_comm_channel_admin CLI tool.
NVIDIAOfficial
Diagnose NVFLARE job failures by analyzing logs and metrics in simulation or production environments.
NVIDIAOfficial
Monitor the progress and status of detached KERMT training or fine-tuning runs.
NVIDIAOfficial
A structured, evidence-gated workflow for optimizing Mixture-of-Experts (MoE) model throughput in Megatron Bridge.
NVIDIAOfficial
Run and operate the DeepStream Multi-View 3D Tracking reference app, also known as MV3DT. Use when the user asks to set up prerequisites, run shipped MV3DT samples, run Multi-View 3D Tracking on custom synchronized MP4 datasets, import camera calibration, delegate missing calibration to AutoMagicCalib, inspect OSD or BEV visualization, consume MV3DT Kafka metadata, or clean up MV3DT run state in the DeepStream MV3DT app directory.
NVIDIAOfficial
Configure and tune Mixture-of-Experts (MoE) communication overlap in Megatron Bridge for improved training throughput.
NVIDIAOfficial
Provides guidance for using DOCA Pipeline Language (DPL) Developer Tools to inspect, debug, and manage DPL programs on BlueField DPUs.