skills.rest
flagos-ai
flagos-aiOFFICIAL
Explore all skills by this author →
WebsiteGitHub

kernelgen-flagos

Automate GPU kernel generation and optimization across FlagGems, vLLM, and generic repositories.

18|12|Updated Mar 3, 2026
Search Tags:#mcp#gpu#triton#vllm#kernelgen#flaggems
One-click install
npx skills add https://github.com/flagos-ai/skills --skill kernelgen-flagos
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: kernelgen-flagos
Source: https://github.com/flagos-ai/skills/tree/main/skills/kernelgen-flagos
Command: npx skills add https://github.com/flagos-ai/skills --skill kernelgen-flagos

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

The Skill bundles nine distinct sub-skills into a single self-contained package, enabling users to generate, optimize, and specialize GPU kernel operators without juggling multiple tools or repositories.

Core Features & Use Cases

  • kernelgen-generate: Generate GPU kernels for Python/Triton repos via MCP.
  • kernelgen-generate-for-flaggems: FlagGems-specific generation workflow (e.g., handling pointwise_dynamic ops and registered operator configs).
  • kernelgen-generate-for-vllm: vLLM-specific generation workflow (e.g., SPDX headers, CustomOp registration).
  • kernelgen-optimize: General Triton kernel optimization via MCP iterative loop.
  • kernelgen-optimize-for-flaggems: FlagGems-specific optimization modes (built-in/external/experimental operators).
  • kernelgen-optimize-for-vllm: vLLM-specific optimization including CustomOp integration.
  • kernelgen-specialize: Platform specialization for Triton operators (e.g., migrating to Ascend NPU).
  • kernelgen-specialize-for-flaggems: FlagGems-integrated platform specialization.
  • kernelgen-submit-feedback: Submit bug reports via GitHub or email.
  • kernelgen-mcp-setup: Check and auto-configure the kernelgen server (MCP).

Quick Start

Use the kernelgen-flagos skill to generate or optimize a GPU kernel by invoking /kernelgen-flagos with the operator name.

Frequently Asked Questions about kernelgen-flagos

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a Triton GPU kernel for a vLLM repository?▼

Yes, you can optimize FlagGems operators using the built-in, external, or experimental modes provided by the kernelgen-flagos skill, which applies an iterative MCP loop to automatically refine and benchmark the Triton kernels.

Can I specialize Triton operators for the Ascend NPU platform?▼

The kernelgen-flagos skill requires the kernelgen MCP server to be configured. It includes a setup sub-skill that automatically checks your environment and configures the MCP server to coordinate code generation and benchmarking tasks.

What is the best way to automate end-to-end GPU kernel development?▼

The best way to automate end-to-end GPU kernel development is using a unified MCP-based workflow that coordinates code generation, environment checks, testing, benchmarking, and feedback submission without juggling multiple tools.

Does generating GPU kernels require manually switching tools for different repositories?▼

No, generating GPU kernels does not require manually switching tools because the kernelgen-flagos skill automatically detects whether you are in a FlagGems, vLLM, or generic repository and dispatches to the corresponding specialized sub-skill.

Related Skills

Explore More Software Engineering Skills→
Srain021Srain021

kestrel-mcp-bootstrap

Onboard MCP agents through structured environment checks and readiness reporting.

Community
Advanced
xchang1121xchang1121
6

triton-ascend-performance-improvement

Apply optimization patterns to Triton Ascend kernels for improved performance.

Community
Advanced
Fridge003Fridge003
9

add-sgl-kernel

Add a CUDA/C++ kernel to the SGL ecosystem with PyTorch integration.

Community
Advanced
AscendAscend
31

vector-triton-ascend-ops-optimizer

Optimizes a Triton operator's performance on Ascend NPU using UB-aware kernel tuning and profiling workflows.

Official
Advanced