What is pull_style_cutedsl_megakernel?

A vendored SM90 FP8 MegaMoE kernel source tree for NVIDIA Hopper GPUs, kept as a separate backend from the SM100 tree for Blackwell GPUs. Vendored code is copied into a project at a specific upstream revision and is maintained there as its own drop.

Use it when updating or working with the SM90 CuTeDSL MegaMoE kernel source, its runtime, or the path setup that selects the Hopper backend.

It prevents incompatible kernel trees and colliding module names from being loaded together. Independent revisions also avoid accidentally mixing files between the Hopper and Blackwell backends.

pull_style_cutedsl_megakernel — Skill for Claude Code and Codex by fla

About the project

FlashInfer is a library and kernel generator that supplies GPU operations used to run large language model inference, including attention, matrix multiplication, and mixture-of-experts computations. It helps engineers build and optimize LLM serving systems across supported GPU hardware and backend implementations. Its catalogue add-ons provide skills and instructions for working with FlashInfer.

flashinfer-ai/flashinfer · 6,370 stars · on GitHub · flashinfer.ai

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent

npx skills add flashinfer-ai/flashinfer --skill pull_style_cutedsl_megakernel

Clone the repo

git clone --depth 1 https://github.com/flashinfer-ai/flashinfer

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/flashinfer-ai/flashinfer/pull_style_cutedsl_megakernel/github.svg)](https://agentmods.dev/skills/flashinfer-ai/flashinfer/pull_style_cutedsl_megakernel)

Your own site

<a href="https://agentmods.dev/skills/flashinfer-ai/flashinfer/pull_style_cutedsl_megakernel"><img src="https://agentmods.dev/badge/skills/flashinfer-ai/flashinfer/pull_style_cutedsl_megakernel/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/flashinfer-ai/flashinfer/pull_style_cutedsl_megakernel"><img src="https://agentmods.dev/badge/skills/flashinfer-ai/flashinfer/pull_style_cutedsl_megakernel.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 0 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,108 The whole file, excluding the scripts and references it only reads on demand.

Security scan C 1 finding. A grade says what 26 rules found in the file — not that it is safe. Third-party audits

NVIDIA SkillSpector warn 7 Sept 2026

SkillSpector: 2 findings, up to high

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

high Tool Misuse · line 58
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
high Tool Misuse · line 61
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.

How audits are shown

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00000	$0.01108
Opus 5	$0.00000	$0.00554
Sonnet 5	$0.00000	$0.00222
Haiku 4.5	$0.00000	$0.00111

Measured 11d ago against content hash d4fd7d210cdf, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade C, and why

pull_style_cutedsl_megakernel scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.

The scan reads SKILL.md. This mod also ships 59 executable files (__init__.py, shim/__init__.py, shim/_paths.py, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

rm -rf flashinfer/moe_ep/kernel_src/sm90/pull_style_cutedsl_megakernel/src/{common,src,moe_nvfp4_swapab,moe_hopper_fp8}

flashinfer/moe_ep/kernel_src/sm90/pull_style_cutedsl_megakernel/SKILL.md · 73 lines

How it starts

The opening of the file, as written. The whole thing — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Updating the SM90 (Hopper) CuTeDSL MegaMoE kernel src

Provenance

This tree vendors the kernel team's SM90 FP8 MegaMoE drop — a fork of the same kernel repo that kernel_src/cutedsl_megamoe vendors (Bangyu's SM100 tree). The SM90 work (Vincent's hopper_megamoe branch) moved the shared runtime forward, so this tree duplicates common/, src/, and moe_nvfp4_swapab/ at its own revision instead of sharing the SM100 copies. The two trees are separate backends:

top-level module names collide (common, src, moe_nvfp4_swapab), so only one tree can be active per process — shim/_paths.bootstrap_paths raises if the sibling tree's modules are already imported. A process runs on either Hopper or Blackwell, never both, so this is not a practical limit.
drops are updated independently; never "sync" shared files across the trees.

Current drop: kernel repo commit 1275b8b ("Merge branch 'vincent/hopper_megamoe' into 'main'", 2026-07).

Layout

kernel_src/sm90/pull_style_cutedsl_megakernel/
├── src/                    ← VERBATIM kernel-team drop; NEVER edit or add files here
│   ├── common/             ← shared constants/host utils (SM90-drop revision)
│   ├── src/                ← CuTeDSL core src (bootstrap, dispatch, sym_buffer, token_comm, …)
│   ├── moe_nvfp4_swapab/   ← NVFP4 package (hopper_fp8 reuses its runner_common,
│   │                          fc1_fc2_fuse_sched, topk_reduce, custom_ext, moe_utils)
│   └── moe_hopper_fp8/     ← SM90 FP8 kernel implementation
│       (benchmark_data/ is excluded from the copy — data blobs only)
├── __init__.py             ← public API for moe_ep; talks ONLY to shim/ (our code)
├── shim/                   ← thin adapters over src/ (our code) — ALL adaptation lives here
│   ├── _paths.py           ← adds sibling src/ to sys.path + sibling-tree exclusivity guard
│   ├── comm.py             ← dist/NVSHMEM bootstrap, sym heap, launch-cache state
│   ├── hopper_fp8.py       ← SM90 FP8 frontend (config, symm buffer, compute entry)
│   └── kernel_helpers.py   ← lazy re-export point for raw-kernel helpers/reference
├── SKILL.md                ← this file (drop-update workflow)
└── TUNING.md               ← measured perf vs the kernel drop's reference sweep,
                              benchmark methodology, knob surface, next levers

Read the full file on GitHub · 73 lines

Files

What ships with it

60 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

11d ago First seen · 73 lines · 0 tokens per session scan C d4fd7d210cdf

Subscribe to this mod's changes

pull_style_cutedsl_megakernel is a skill published in the GitHub repository flashinfer-ai/flashinfer (6,370 stars, last pushed today), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,108 tokens. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

GPU PyTorch cuda jit large-large-models llm-inference

gpu-server-setup

Prepare a Linux server with NVIDIA GPUs for neural-network and LLM workloads, or diagnose one that misbehaves. Use when the user asks to set up or prepare a GPU server, install NVIDIA drivers or CUDA on Debian/Ubuntu, wire Docker to GPUs (NVIDIA Container Toolkit), deploy vLLM / Infinity / OpenWebUI / Ollama /…

EvilFreelancer/gpu-server-setup · 167 tokens

cuopt-skill-evolution

After solving a non-trivial problem, detect generalizable learnings and propose skill updates. Always active — applies to every interaction.

NVIDIA/cuopt · 33 tokens

cuopt-developer

Modify, build, test, debug, and contribute to NVIDIA cuOpt (C++/CUDA, Python, server, CI). Use for solver internals, PRs, DCO, and code conventions.

NVIDIA/cuopt · 47 tokens

cuopt-user-rules

Base rules for end users calling NVIDIA cuOpt (routing/LP/MILP/QP/install/server). Not for cuOpt internals — use cuopt-developer for those.

NVIDIA/cuopt · 42 tokens

perf-optimize

Launch and operate this repo's perf-optimize workflow, which iteratively APPLIES TensorRT-LLM serving optimizations — baseline benchmark at one concurrency or a Pareto curve of them (tok/s/user vs tok/s/gpu), analytical SOL projection on by default (via the internal-perf-sol-analysis skill) sizing the headroom the…

NVIDIA/TensorRT-LLM · 224 tokens

perf-analyze

Launch and operate this repo's perf-analyze workflow, which DIAGNOSES a TensorRT-LLM serving deployment without applying changes — benchmark at one concurrency or a Pareto curve of them (tok/s/user vs tok/s/gpu), analytical SOL projection on by default (via the internal-perf-sol-analysis skill), nsys + ncu per-kernel…

NVIDIA/TensorRT-LLM · 159 tokens

pull_style_cutedsl_megakernel