tilekernels-gpu-kernels — Claude Code skill
This skill guides agents through TileKernels GPU kernels for MoE, quantization and other LLM operations. Use it when writing or checking TileLang code.
npx skills add reason-machines/trending-skills --skill tilekernels-gpu-kernels -a claude-codeWhat tilekernels-gpu-kernels does
The skill focuses on TileKernels, a GPU kernel library for LLM workloads. It covers MoE routing, FP8 and FP4 quantization, batched transpose, engram gating and Manifold HyperConnection. It also points to PyTorch-facing wrappers and reference implementations used to validate kernel results.
The covered layout includes moe, quant, transpose, engram, mhc, modeling, torch and testing. An agent can work with top-k expert selection, token-to-expert mapping, token or block quantization, Sinkhorn normalization and pytest checks.
The stated setup requires Python 3.10+, PyTorch 2.10+, TileLang 0.1.9+ and CUDA Toolkit 13.1+. The kernels target NVIDIA GPUs with SM90 or SM100 architecture, so other GPU targets fall outside the stated setup.
What it helps with
- choose kernels for MoE routing
- quantize activations with FP8 or FP4
- transpose batched GPU tensors
- compare a kernel with its PyTorch reference
- run TileKernels tests and benchmarks
Example prompts for the agent
- Use TileKernels to design an MoE route with top-k expert selection.
- Write an FP8 activation quantization path in TileLang and compare it with the PyTorch reference.
- Inspect the
transposemodule and run itspytestchecks.
How to install tilekernels-gpu-kernels
1Skills CLI (Vercel)
npx skills add reason-machines/trending-skills --skill tilekernels-gpu-kernels -a claude-code2Manually: put the files in place
~/.claude/skills/tilekernels-gpu-kernels/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/tilekernels-gpu-kernels/
1Skills CLI (Vercel)
npx skills add reason-machines/trending-skills --skill tilekernels-gpu-kernels -a codex2Manually: put the files in place
.agents/skills/tilekernels-gpu-kernels/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add reason-machines/trending-skills --skill tilekernels-gpu-kernels -a cursor2Manually: put the files in place
.cursor/skills/tilekernels-gpu-kernels/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install reason-machines/trending-skills tilekernels-gpu-kernels2Manually: put the files in place
.github/skills/tilekernels-gpu-kernels/SKILL.md1Skills CLI (Vercel)
npx skills add reason-machines/trending-skills --skill tilekernels-gpu-kernels -a gemini-cli2Platform CLI
gemini skills install https://github.com/reason-machines/trending-skills --path skills/tilekernels-gpu-kernelsManually: put the files in place
.devin/skills/tilekernels-gpu-kernels/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add reason-machines/trending-skills --skill tilekernels-gpu-kernels -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about tilekernels-gpu-kernels
How do I install tilekernels-gpu-kernels in Claude Code?
Run npx skills add reason-machines/trending-skills --skill tilekernels-gpu-kernels -a claude-code, then use ~/.claude/skills/tilekernels-gpu-kernels/SKILL.md.
What does tilekernels-gpu-kernels do?
It guides the agent through TileKernels GPU operations for LLMs, including MoE routing, quantization, transpose, engram gating and Manifold HyperConnection.
What is required to run TileKernels?
The stated requirements are Python 3.10+, PyTorch 2.10+, TileLang 0.1.9+, CUDA Toolkit 13.1+ and an NVIDIA GPU with SM90 or SM100 architecture.
