GPU Test

Identify GPUs, run bounded compute or model inference benchmarks, and compare saved before/after results on NVIDIA, AMD, Intel, or Apple systems using supported installed runtimes. Use for GPU performance and basic numerical checks; includes a separate explicit workflow for qualified Linux NVIDIA pre-driver unlock trials.

12 starsTesting & QA

Install

skills CLI (Claude Code, Cursor, Codex and more)

npx skills add xero00000/gpu-test --skill gpu-test

Manual install

git clone --depth 1 https://github.com/xero00000/gpu-test.git /tmp/gpu-test
mkdir -p ~/.claude/skills
cp -r /tmp/gpu-test/. ~/.claude/skills/gpu-test

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# GPU Test

Produce measured results for the user's actual GPU and workload. Discover hardware first; select an installed runtime that supports it. An unsupported backend is a reported limitation, never a CPU benchmark presented as GPU performance.

## Run a short compute test

Use the bundled Python helper (paths below are relative to this skill):

```bash
python3 scripts/gpu_test.py inventory --output /absolute/new-run/inventory.json
python3 scripts/gpu_test.py bench --output /absolute/new-run/compute
```

Inventory works without PyTorch. The benchmark uses an existing PyTorch installation: CUDA for NVIDIA, ROCm through PyTorch's `cuda` API for AMD, XPU for Intel, or MPS for Apple. Choose another installed interpreter when appropriate. Do not install large runtime packages or change drivers merely to make the helper work; inspect existing tools and use a suitable native benchmark when PyTorch is unavailable. Consult current vendor documentation for unfamiliar commands or support claims.

Defaults: all devices of the first available backend, sequential workers, 4096-square GEMMs, three warmups, five samples of three operations, and a 60-second worker timeout. Explicit selection:

```bash
python3 scripts/gpu_test.py bench --backend cuda --devices 1 --size 4096 --output /absolute/new-run/card-1
python3 scripts/gpu_test.py compare /absolute/before/summary.json /absolute/after/summary.json --output /absolute/comparison.json
```

Read the full skill on GitHub

Details

Skill name: gpu-test
Repository: xero00000/gpu-test
License: MIT
Repo updated: 24 October 2026

FAQ

What does the GPU Test skill do?

How do I install the GPU Test skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add xero00000/gpu-test --skill gpu-test. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is GPU Test popular?

It lives in xero00000/gpu-test, which has 12 GitHub stars and 0 forks.

Related skills

Trip Planner

TianhaoWu66/trip-planner

Plan scenic multi-day domestic trips (China) end-to-end and deliver a detailed, verified travel guide as Markdown + PDF. Use when the user asks for a travel itinerary/攻略 (e.g. 帮我制定北疆旅游攻略, 做个云南行程, 计划一次川西自驾). Workflow: first confirm travel dates and the scenic-spot list (from the user or researched via 携程攻略 sights page + 小红书/马蜂窝), source long-haul transport from 12306 or airlines, plan inter-city routes with 高德地图, verify every ticket/reservation with a cited source (never fabricate), compare real hotel prices on Ctrip and Fliggy and verify restaurant ratings on Dianping (the customer must log in

Testing & QA

Serial Protocol Assistant

10walnut/serial-protocol-tester-skill

Convert vendor serial communication specifications into validated, single-language serial_protocol.v1 JSON that drives a functional test console. Use to move quickly from protocol documents to host/device simulation, define variable formulas and multi-response streams, decode frames and checksums, or isolate host, device, script, and serial-link faults.

Testing & QA

Forge

ikennaokpala/forge

Autonomous quality engineering swarm that forges production-ready code through continuous behavioral verification, exhaustive E2E testing, and self-healing fix loops. Combines DDD+ADR+TDD methodology with BDD/Gherkin specifications, 7 quality gates, defect prediction, chaos testing, and cross-context dependency awareness. Architecture-agnostic - works with monoliths, microservices, modular monoliths, and any bounded-context topology.

Testing & QA

Red Green Mode

hidevinliu/red-green-mode

Autonomous test-fix loop: run an objective verifier (tests, type-check, lint, CI), fix what is red, re-run, and finish only when the verifier exit code is all green. Use when the user wants work driven to green without supervision, including long-run or parallel runs (红绿灯模式 / 跑到全绿 / 把 CI 修到全绿). Not for subjective work with no verifier or one-shot edits; use test-driven-development to write tests and systematic-debugging to root-cause one bug.

Testing & QA

Healthsim

mark64oswald/healthsim-workspace

HealthSim generates realistic synthetic healthcare data for testing EMR systems, claims processing, pharmacy benefits, and analytics. Use for ANY request involving: (1) synthetic patients, clinical data, or medical records, (2) healthcare claims, billing, or adjudication, (3) pharmacy prescriptions, formularies, or drug utilization, (4) HL7v2, FHIR, X12, or NCPDP formatted output, (5) healthcare testing scenarios or sample data generation.

Testing & QA

Does It Work

TsakunovR/does-it-work

Проверка, что продукт реально работает, и защита его качества автотестами. Аудит работающего (в т.ч. навайбкоженного) приложения: найти баги, оценить готовность к проду, выдать баг-репорт с severity. Генерация тестовых фреймворков: API-тесты на Python (pytest + httpx + Pydantic + Allure) и Java (JUnit 5 + REST Assured), WEB/UI-тесты (Playwright, Selenium, Selenide, Page Object); конвертация OpenAPI/Postman/curl/HAR в тесты, негативные кейсы, контракты. Также стабилизация флакающих тестов и ревью качества тестов. Triggers: "проверь мой продукт/приложение", "найди баги", "навайбкодил", "можно ли

Testing & QA

Browse the most popular Claude skills, all testing & qa skills, or write your own with the skills generator.