Write, debug, and optimize Triton and Gluon GPU kernels.
npx skills add https://github.com/LeightonWang/flashinfer --skill triton-skills
Please help me install this Agent Skill. Skill: triton-skills Source: https://github.com/LeightonWang/flashinfer/tree/main/.github/skills/triton-skills Command: npx skills add https://github.com/LeightonWang/flashinfer --skill triton-skills
This Skill helps you write, debug, and optimize Triton and Gluon GPU kernels using local source code, tutorials, production kernel references, and compiler internals.
Ask for help implementing, debugging, or optimizing a Triton or Gluon kernel and include the target operation, tensor shapes, and hardware constraints.
High-intent search queries and answers about installing and using this skill.
To optimize Triton kernels, apply practical patterns for autotuning, persistent kernels, and warp specialization to improve throughput. This Skill provides reference implementations and source code to help tune operations like fused FP8 attention and matmul.
Debugging Triton kernel layout issues and numerical behavior is best done using curated tutorials and production implementation references. This approach helps identify correct API usage and resolve constraints for technologies like TMA and WGMMA.
Yes, you can implement fused MoE matmul and SwiGLU operations using Triton and Gluon. This Skill provides kernel development guidance and production references to build and optimize these specific GPU scenarios.
Triton supports architecture-specific GPU features including TMA, WGMMA, and tcgen05. This Skill helps you match Triton language semantics and layout behavior to effectively utilize these hardware capabilities in your kernels.
Start writing a high-performance Triton kernel by asking for help with your target operation, tensor shapes, and hardware constraints. This Skill leverages local source code and production references to guide the implementation and optimization process.
fla-orgPort Triton kernels to Gluon with explicit tensor layout and memory control.
m0atDevelop, debug, and autotune Triton and Gluon GPU kernels for ML workloads.
xiechyDetect host CPU, GPU, memory, and disk resources into a .claude_resources.json file.
flashinfer-aiExtract GPU kernel schemas from SGLang models into deduplicated Definition JSON files.
