Home / troycheng / cuda-kernel-optimizer · GitHub

cuda-kernel-optimizer SKILL.md skillA

cuda-kernel-optimizer from troycheng/cuda-kernel-optimizer is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned
mdr add troycheng/cuda-kernel-optimizer/cuda-kernel-optimizer@git:20261009.8ed7d7e

Pins git:20261009.8ed7d7e in mdr.lock. Pin the fingerprint instead.

Audit grade
A, 17 of 17 checks passed
Kind
SKILL.md skill
Repository
troycheng/cuda-kernel-optimizer
Path
skills/cuda-kernel-optimizer/SKILL.md
Versions recorded
5
Latest version
git:20261009.8ed7d7e, committed Oct 9, 2026
First version recorded
Aug 28, 2026
Size
9,218 bytes
Repository stars
7
License
MIT

Use when optimizing or diagnosing GPU workload performance with CUDA, CUTLASS, Triton, PyTorch, vLLM or TensorRT-LLM; interpreting NCU, Nsys or PyTorch Profiler evidence; or preparing the workload and correctness checks for such optimization. General coding, skill maintenance and conceptual GPU questions do not require the optimization workflow.

Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.

What the file says

# CUDA Kernel and Workload Optimizer

ChatGPT 负责优化判断：识别瓶颈、提出候选、评估投入产出并决定下一步。工具执行明确请求，完成检查、测量、解析和事实记录；工具不选择方向，不判断 ROI，也不生成下一步计划。

围绕用户指定的目标优化。kernel 测量支持 kernel 层结论；workload 或服务层收益必须由用户的真实测试集、精度校验和成对测量支持。不得擅自编造、下载或替换测试 workload。

## 工作与沟通

以一个能改变判断的工作步骤组织执行。同一已说明步骤中的编辑、构建和必要检查可以连续完成；在开始昂贵实验、改变方向或预计投入明显增加时，简要说明证据缺口、预期收益、成本和继续条件。已有授权内的常规操作直接推进，长任务及时报告实质进展。

根据问题选择所需操作。已有报告可以直接分析，简单工具错误可以直接定位修复；完整性能优化才需要建立 Target 和测量链路。验证做到足以支持当前结论，检查通过且没有新疑点时继续交付。

## 按需读取

只加载当前问题需要的资料。第一次使用或接口不明确时查看脚本 `--help` 和 `references/request_protocol.md`；已经确认的接口可直接复用。

| 任务 | 脚本或资料 |
|---|---|
| 检查测试集、精度校验、driver、环境并冻结 Target | `scripts/readiness.py`；`references/environment_readiness.md` |
| baseline、Experiment、screen、target 和 final audit | `scripts/workload_evaluate.py`；`references/performance_iteration.md` |
| Nsight Compute | `scripts/profile_ncu.py`；`references/ncu_metrics_guide.md` |
| Nsight Systems | `scripts/profile_nsys.py` |
| PyTorch Profiler | `scripts/profile_pytorch.py` |
| 编译产物和 SASS | `scripts/compiler_evidence.py`；`scripts/sass_check.py` |
| 离线知识 | `scripts/knowledge_query.py` |
| 选择或恢复最佳版本 | `scripts/champion.py` |
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint
mdr add troycheng/cuda-kernel-optimizer/cuda-kernel-optimizer@sha256:10f7cba1741660d4

The line at the top of this page pins the label git:20261009.8ed7d7e, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_aktw4cn6jk7rxn6w.svg)](https://markdownregistry.com/a/art_aktw4cn6jk7rxn6w)

1 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 5 versions of this file, between Aug 28, 2026 and Oct 9, 2026, about one every 10 days.

versioncommittedcommitsizeaudit
git:20261009.8ed7d7e latest2026-10-09 8ed7d7e 9,218 BA view · diff
git:20260905.8384f592026-09-05 8384f59 8,838 BA view · diff
git:20260904.89b10162026-09-04 89b1016 16,985 BA view · diff
git:20260831.f6aa10f2026-08-31 f6aa10f 16,724 BA view · diff
git:20260828.8b52a4e2026-08-28 8b52a4e 15,450 BA view

Audit of the latest version

A  17 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (9218 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No link to a raw IP address
  • pass: No script tag

Source

GitHub

troycheng/cuda-kernel-optimizer · 7 stars · license MIT · pushed 2026-11-04 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_aktw4cn6jk7rxn6w
GET https://markdownregistry.com/api/v1/resolve?ref=troycheng/cuda-kernel-optimizer/cuda-kernel-optimizer
GET https://markdownregistry.com/api/v1/blob/10f7cba1741660d442fa16b75cdbcdb6d9eebfb6227c5a9e4917bed1d2c55da4

Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.

Browse by kind, by grade A, or by owner.