sgl-projectOfficial·29 Agent Skills Included

sglang

Serve and optimize large language and diffusion models at scale

Serves large language models and image/video diffusion models with high throughput on GPUs, TPUs, and NPUs. Eliminates slow inference, memory waste, and manual tuning through prefix caching, quantization, and multi-GPU parallelism. Includes agent skills for kernel development, CI management, performance profiling, and production incident debugging.

33,028|8,428

↗

npx skills add sgl-project/sglang --all -g -y

Available:

Guides the AI agent through SGLang's diffusion subsystem architecture, model-adding workflow, and performance skills, plus rules for writing accurate Mintlify documentation pages.

All Skills in This Repository (29)

Pure Emerald Level Indicators

📦 In Repo

sgl-project

add-new-diffusion-model

Guide adding new diffusion models to the SGLang framework.

diffusion-optimal-perf

Configure SGLang-Diffusion CLI flags and environment variables for inference speedups.

diffusion-kernel

Integrate custom JIT CUDA and Triton kernels to accelerate diffusion model inference.

write-sglang-test

Create SGLang CI and unit tests using CustomTestCase and server fixtures.

add-jit-kernel

Guide adding lightweight JIT CUDA kernels to the sglang framework.

add-sgl-kernel

Guide adding AOT CUDA/C++ kernels to the sgl-kernel library.

SGLang Bisect CI Regression

Bisect CI run history to identify commits introducing test regressions.

sglang-diffusion-modelopt-quant

Quantize diffusion DiT models with NVIDIA ModelOpt and validate FP8 or NVFP4 checkpoints in SGLang Diffusion.

sglang-diffusion-benchmark-profile

Benchmarks denoise latency and profiles diffusion hotspots in SGLang multimodal generation.

sglang-diffusion-add-model

Implements new diffusion models and Diffusers pipelines in SGLang multimodal_gen.

sglang-diffusion-performance

Selects optimal SGLang Diffusion performance flags for a given model, GPU, and VRAM budget.

clean-startup-log

Detect and fix noisy warnings and duplicate prints in SGLang server startup logs.

Official

Advanced

Frequently Asked Questions

FAQPage Schema

How to install SGLang skills?▼

Run `npx skills add sgl-project/sglang --all -g -y` in your terminal to install all skills in this suite globally.

What is SGLang used for?▼

SGLang is a high-performance serving framework for large language models and multimodal diffusion models, delivering low-latency inference from a single GPU to large distributed clusters.

How do I debug a CUDA crash in SGLang?▼

The debug-cuda-crash skill uses kernel API logging to capture input tensors before a crash, helping you trace illegal memory access, NaN values, and shape mismatches.

Can AI agents help tune SGLang performance?▼

Yes. Skills like sglang-diffusion-performance and llm-torch-profiler-analysis guide agents to pick optimal flags, profile traces, and map bottlenecks to existing fast paths.

Do these skills work with Claude Code and other agents?▼

Yes. All skills follow the universal SKILL.md standard and run in Claude Code, OpenClaw, Cursor, and compatible AI coding agents.