FrostyLeaves

Nsight Graphics Analyzer — Development skill for Claude Code

Development community

CLI wrapper and Claude Code / Codex skill for NVIDIA Nsight Graphics 2026.1+ — capture GPU frames, export GPU Trace, and drill into bottlenecks via compact JSON.

How to install Nsight Graphics Analyzer

This entry records only its repository, not the path inside it, so there is no exact command to give. Open FrostyLeaves/nsight-graphics-analyzer and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Nsight Graphics Analyzer does

CLI wrapper and Claude Code / Codex skill for NVIDIA Nsight Graphics 2026.1+ — capture GPU frames, export GPU Trace, and drill into bottlenecks via compact JSON.

Alternatives in Development

  • Persona Analyzer — Persona 分析 Prompt 24k ★
  • Rust Analyzer Lsp — Rust /plugin install rust-analyzer-lsp@claude-plugins-official 14k ★
  • Setup Video Vision — Interactive setup wizard for claude-video-vision — configure backend, whisper, frames, and verify dependencies 1.3k ★

README

nsight-graphics-analyzer

**English** | [简体中文](./README.zh-CN.md)

Nsight Graphics Analyzer is a Windows-only skill and CLI wrapper for **NVIDIA Nsight Graphics 2026.1+**. It helps Claude Code, Codex, or a human at the terminal capture GPU frames, export Nsight GPU Trace data, and inspect the result through small JSON files instead of multi-hundred-MB TSV artifacts.

The Python code uses only the standard library. No `pip install` is required.

[![ci](https://github.com/FrostyLeaves/nsight-graphics-analyzer/actions/workflows/ci.yml/badge.svg)](https://github.com/FrostyLeaves/nsight-graphics-analyzer/actions/workflows/ci.yml) [![license: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE) ![python](https://img.shields.io/badge/python-3.10%2B-blue) ![platform](https://img.shields.io/badge/platform-Windows%2010%2F11-lightgrey)

Requirements

Requirement Notes
Windows 10/11 Required. The wrapper uses Windows process and path APIs.
NVIDIA Nsight Graphics 2026.1+ Install separately from NVIDIA Developer.
Python 3.10+ Standard library only.
NVIDIA GPU and supported driver Required for Nsight capture and replay workflows.

Run the environment check first:

python plugins\nsight-graphics-analyzer\skills\nsight-graphics-analyzer\scripts\nsight.py doctor

If `ngfx_install` is `null`, install Nsight Graphics before trying to capture.

Quick Start

Capture a short GPU Trace and generate the three compact JSON artifacts:

python plugins\nsight-graphics-analyzer\skills\nsight-graphics-analyzer\scripts\nsight.py gputrace-capture `
  --exe "C:\Path\To\YourApp.exe" --wd "C:\Path\To" `
  --start-after-ms 5000 --max-duration-ms 1000 `
  --architecture Ada --metric-set-name "Throughput Metrics" `
  --time-every-action `
  --out "C:\Path\To\Captures\sample.ngfx-gputrace"

The wrapper writes these files next to the trace:

| Artifact | Purpose |