What is cutedsl_megamoe?

A set of project instructions for updating the source and adapters of a CuTeDSL MegaMoE kernel, a GPU program used by machine-learning models with many expert parts.

Use it when importing a new kernel-source version, updating the adapter code, or checking the layout of the MegaMoE kernel files.

It clarifies which files come from the kernel team and must stay unchanged, and where project-specific adaptations belong.

cutedsl_megamoe — Skill for Claude Code and Codex by flashinfer-ai

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 cutedsl_megamoe

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/cutedsl_megamoe/github.svg)](https://agentmods.dev/skills/flashinfer-ai/flashinfer/cutedsl_megamoe)

Your own site

<a href="https://agentmods.dev/skills/flashinfer-ai/flashinfer/cutedsl_megamoe"><img src="https://agentmods.dev/badge/skills/flashinfer-ai/flashinfer/cutedsl_megamoe/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/cutedsl_megamoe"><img src="https://agentmods.dev/badge/skills/flashinfer-ai/flashinfer/cutedsl_megamoe.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 2,509 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: 1 finding, 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 67
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	$0.02509
Opus 5	$0.00000	$0.01255
Sonnet 5	$0.00000	$0.00502
Haiku 4.5	$0.00000	$0.00251

Measured 11d ago against content hash 4a4f8d3f165a, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade C, and why

cutedsl_megamoe 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/__main__.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/cutedsl_megamoe/src/{common,src,moe_bf16_glu,moe_mxfp8_glu,moe_nvfp4_swapab}

Origin

Copies of this mod

1 near-identical copy found in the catalogue:

cadence-tldr — 88% identical, 142 lines differ

flashinfer/moe_ep/kernel_src/cutedsl_megamoe/SKILL.md · 138 lines

How it starts

The opening of the file, as written. The whole thing — 138 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Updating the CuTeDSL MegaMoE kernel src

Layout

kernel_src/cutedsl_megamoe/
├── src/                    ← VERBATIM kernel-team drop; NEVER edit or add files here
│   ├── common/
│   ├── src/                ← CuTeDSL core src (bootstrap, dispatch, sym_buffer, …)
│   ├── moe_mxfp8_glu/      ← MXFP8 kernel implementation
│   ├── moe_bf16_glu/       ← BF16 kernel implementation
│   └── moe_nvfp4_swapab/   ← NVFP4 kernel implementation
├── __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 (bootstrap_paths); shim glue
│   ├── comm.py             ← dist bootstrap, sym heap, compile state, resolve_gate_up_clamp
│   ├── nvfp4.py            ← NVFP4 frontend + symm-buffer/launch wrappers (self-contained)
│   ├── mxfp8.py            ← MXFP8 frontend + symm-buffer/launch wrappers (self-contained)
│   ├── bf16.py             ← BF16 frontend + symm-buffer/launch wrappers (self-contained)
│   ├── kernel_helpers.py   ← SINGLE re-export point for raw-kernel helpers/constants/
│   │                          reference the FI backend + tests need (drop-audit point)
│   ├── tuner.py            ← kernel tuning knobs (tactic enumeration + config apply);
│   │                          mirrors tester/solvers/inference_solver knob taxonomy
│   ├── autotune.py         ← online (warmup-time) COLLECTIVE knob autotuning:
│   │                          times a curated candidate set on the live problem,
│   │                          all-reduces (MAX) across ranks, applies the winner
│   │                          (backends trigger it via config knobs="auto")
│   └── correctness.py      ← standalone NVFP4 smoke runner (not used by moe_ep)
├── SKILL.md                ← this file (drop-update workflow)
└── TUNING.md               ← tuning surface + measured profiles + benchmarking
                               methodology (read before re-tuning or comparing
                               against deep_gemm / the kernel-repo tester)

Read the full file on GitHub · 138 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 · 138 lines · 0 tokens per session scan C 4a4f8d3f165a

Subscribe to this mod's changes

cutedsl_megamoe 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 2,509 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.

C and C++Data and AI GPU Languages PyTorch cuda jit large-large-models llm-inference

cutedsl_megamoe

Getting it into your agent

Wrote this? Show the measurements

What it costs to keep this loaded

Grade C, and why

Recursive force deletehighDestructive command

Copies of this mod

How it starts

Updating the CuTeDSL MegaMoE kernel src

Layout

What ships with it

What this file has done since we first saw it

Other skills, from other repositories

cuopt-developer

transform-model

ax-cpp-ai

ax-cpp-gen

ax-cpp-llm

ax-cpp-signature