AMD Skills: Empowering AI Agents with AMD's Optimized Software Stack

Introduction

AMD Skills is an official catalog of AI agent skills provided by AMD. It aims to empower AI agents with AMD's optimized software stack, offering a collection of knowledge, scripts, and conventions tailored for AMD hardware and software. These skills follow the standardized Agent Skills format and are designed to interoperate seamlessly with popular coding agents such as Cursor, Claude Code, OpenAI Codex, and Gemini CLI. By leveraging AMD Skills, developers can streamline workflows for tasks like local AI use, LLM inference on AMD GPUs, and performance optimization.

Installation

Installing AMD Skills is straightforward, primarily using the skills CLI via npx. This method requires Node.js.

To add the AMD Skills catalog:

npx skills add amd/skills

This command will prompt you to select a specific skill and an installation destination. If you wish to install a particular skill for specific agents, use the --skill and --agent flags:

npx skills add amd/skills --skill local-ai-use --agent claude-code

You can also browse all available skills before installation:

npx skills add amd/skills --list

For those who prefer a manual approach, you can clone the repository and copy or symlink the desired skill folders into your agent's skills directory. Each agent automatically discovers SKILL.md files.

git clone https://github.com/amd/skills.git amd-skills
cp -r amd-skills/skills/local-ai-use <agent-skills-dir>/

Common agent skills directories include:

• Cursor: ~/.cursor/skills/ or .cursor/skills/
• Claude Code: ~/.claude/skills/ or .claude/skills/
• Codex: $HOME/.agents/skills or $REPO_ROOT/.agents/skills

Examples

Once a skill is installed, you can reference it in plain language when interacting with your AI agent. The agent is often capable of picking the correct skill based on its description, making explicit invocation a fallback rather than a primary requirement.

Here are some examples of how you might use AMD Skills:

• "Use AMD Skills to learn how to generate images locally instead of burning cloud tokens."
• "Use AMD Skills to deploy this LLM for inference on my AMD Instinct GPUs."

For more detailed, step-by-step guides demonstrating skills in action, refer to the walkthroughs provided within the repository.

Why use AMD Skills?

AMD Skills offers a distinct advantage over traditional documentation by providing opinionated, actionable workflows. While documentation describes an API surface, listing every flag and option neutrally, a skill encodes a specific, optimized path. It captures the decisions a senior AMD engineer would make without extensive thought, such as which flags, container image, gfx target, or environment variables to use, and in what specific order.

This approach ensures the consistent application of best practices across different teams and repositories. Skills are particularly valuable for repeated, complex workflows that are common within the AMD software stack, helping agents apply expert knowledge reliably and efficiently.

Links

• GitHub Repository: amd/skills
• License: MIT License