amd-skills Claude Code Plugin
Claude Code PluginClaude CodeDevelopmentAI Agents & AI App DevelopmentClaude Code Customization & WorkflowCode Quality, Review & TestingThe explanation below is AI-generated. Please verify it against the sources.
amd-skills is a plugin that bundles a catalog of AMD-authored Agent Skills for use with coding agents such as Claude Code, Cursor, OpenAI Codex, and Gemini CLI. According to the README, the skills span client-side Ryzen AI workflows (routing image and audio tasks to a local AI server), server-side AMD Instinct GPU workflows (deploying and benchmarking LLM inference with vLLM), throughput optimization on Instinct GPUs via Hyperloom, and performance analysis of GPU kernels and PyTorch traces via tools such as Magpie and TraceLens. Each skill is a self-contained folder with a SKILL.md file that an agent loads when the task matches, plus optional scripts and reference material. The catalog is described as federated, meaning individual skills are maintained in their owning AMD product repositories and pulled into this repository automatically. Some listed skills, such as rocm-doctor and hrr-replay-analysis, are marked as planned rather than currently available.
What you can do with amd-skills
- Integrate local AI capabilities into existing cloud-based LLM applications for offline use, added privacy, and lower API costs
- Set up a Lemonade model router that dispatches requests based on content or required capability
- Deploy and benchmark LLM inference on AMD Instinct GPUs using vLLM, including hardware detection and model-fit validation
- Evaluate GPU kernel correctness and performance and compare kernel implementations using Magpie
- Orchestrate PyTorch profiler trace analysis with TraceLens to produce prioritized performance reports
Sources
History of amd-skills
- Claude Code Plugin Updated the official description of amd-skills
- Claude Code Plugin Added amd-skills