vllm-omni-skills plugin
Seventeen skills for vLLM-Omni covering setup, serving, image, video and speech generation, tuning and PR review.
- 89 ★ on the repository
- Last commit
- Apache-2.0
Minimal access
Why this level
- The skills suggest install and launch commands the agent may run on your machine
Install
- Works in
Manual install
/plugin marketplace add hsliuustc0106/vllm-omni-skillsAfter adding the marketplace, a skill installs with a command like `/plugin install vllm-omni-review@vllm-omni-skills`.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skills suggest install and launch commands the agent may run on your machine
This is third-party code. Review the repository files before installing.
What it does
The collection teaches an agent to work with vLLM-Omni, a framework for serving models with text, image, video and audio. The index has 17 skills: installation and GPU prerequisites, an OpenAI-compatible API client, launching servers, hardware backends (CUDA, ROCm, NPU, XPU), image, video and speech generation, distributed inference, performance tuning, quantization, PR review and release note drafting. Skills activate by context: a question about installation triggers `vllm-omni-setup`, and a request to generate an image brings in `vllm-omni-image-gen`. Each skill is packaged as its own Claude Code plugin, so you can take only the one you need, such as `vllm-omni-review`.
Who it is for. Engineers who deploy vLLM-Omni on their own GPUs and want the agent to know the framework and its models.
Good fit when
- Installing vLLM-Omni stalls on drivers and environment, and the agent needs step-by-step guidance.
- A server for image or speech generation is configured for a specific model, such as FLUX or Qwen3-TTS.
- A distributed run on four GPUs needs role layout and Ray configuration.
- A pull request to the vLLM-Omni project is checked against the review checklist.
Not a fit when
- Work uses plain vLLM for text: the skills target the multimodal framework.
- No GPU or server access is available: the advice covers launching and tuning inference.
- Project documentation itself is needed: the skills condense practice and do not replace official pages.
Example request
How do I install vllm-omni on a server with four GPUs and serve the Qwen-Image model for image generation?Limitations
The last commit is dated July 2026, so model descriptions and build parameters may lag behind fresh vLLM-Omni releases. In Claude Code skills install as plugins one by one, while Cursor and Codex need the `skills/` directory copied by hand. The descriptions summarize framework capabilities, and checking commands on your hardware is up to you: CUDA, driver and model versions affect results. Heavy video generation models need a GPU with plenty of memory.
How to disable. In Claude Code remove the plugins with `/plugin uninstall`, and for Cursor and Codex delete the copied skill folders.
FAQ
How do I install it in Claude Code?
Add the repository as a marketplace with `/plugin marketplace add hsliuustc0106/vllm-omni-skills`, then install skills one at a time in `plugin@marketplace` form, for example `vllm-omni-review@vllm-omni-skills`.
How do I install it in Cursor and Codex?
For Cursor copy `skills/` into the project's `.cursor/skills/`, and for Codex run `cp -r skills/* ~/.codex/skills/`.
Which models are covered?
Images name FLUX, SD3, Qwen-Image and BAGEL, video names Wan2.2, audio names Qwen3-TTS, MiMo-Audio and Stable-Audio, and end-to-end models name Qwen-Omni.
How do skills turn on?
On their own, by the meaning of the request. An installation question activates `vllm-omni-setup`, and a distributed run request brings in `vllm-omni-distributed`.
Is there a skill for contributing to the project?
Yes, `vllm-omni-contrib` covers adding new models and the development workflow, and `vllm-omni-recipe` helps write deployment guides.
Discussion
No questions or reviews yet.
Related
Pluginspstack skill catalog ported to OMP: a task router, TDD, code review and subagent playbooks, each invoked by its own command.
A set of 21 skills for Odoo addon development, OCA migration, reports, web security and architecture review.
73 skills for a delivery pipeline: plan, code, review and QA, ending in a ready, labeled PR in your own repository.
A plugin of 24 Claude Code skills for AI apps on Next.js, AI SDK 7, shadcn/ui and pgvector, with in-browser UI checks
15 skills for Claude Code and Codex: Laravel, Docker, UI, debugging, tests and docs in four ready plugin bundles
Claude Code Skills and Agents
Claude Code Skills & Agents
A Claude Code plugin bundling skills and sub-agents: deep research, image generation, GitHub work, code review and model switching
README in short
Gives an index of 17 skills with descriptions, installation for Cursor, Claude Code and Codex and example questions that activate skills.