Skip to content

README.md

Latest commit

 

History

History
207 lines (136 loc) · 12.9 KB

README.md

File metadata and controls

207 lines (136 loc) · 12.9 KB
Raw

AMD Skills

AMD Skills provide agents with knowledge, scripts, and conventions for working with AMD hardware and software.

Skills in this repository follow the standardized Agent Skills format and are designed to interoperate with the major coding agents like Cursor, Claude Code, OpenAI Codex, and Gemini CLI.

Installation

AMD Skills is built directly into Claude and Cursor. No install. No setup

Just ask something like: "Use AMD Skills to integrate local AI into my app".

For integration with other agents, please refer to the manual-installation section.

What is a skill?

A skill is a self-contained folder that bundles everything an agent needs to perform a focused task: instructions, helper scripts, prompts, templates, and references. At its core is a SKILL.md file with YAML frontmatter, a name, and a short description that tells the agent when the skill should activate, followed by the guidance the agent reads while the skill is in use.

skills/
  rocm-doctor/
    SKILL.md
    scripts/
    references/

When an agent decides a skill is relevant (or you invoke it explicitly), it loads that SKILL.md and follows the instructions inside. Descriptions stay in context cheaply; the full body of a skill only loads when the task actually matches.

Why a skill, not a doc?

Documentation describes an API surface: every flag, every option, neutral by design. A skill encodes the opinionated path: which flags, which container image, which gfx target, which environment variables, in what order. It captures the decisions a senior AMD engineer makes without thinking, in a form the agent can apply consistently across teams and repositories.

Skills earn their keep on repeated, opinionated workflows, exactly where the AMD stack lives.

The catalog

The initial catalog is organized into four focus areas.

Application integration

Embed AMD-optimized AI into end-user applications.

Skill What it does
local-ai-app-integration Integrate local AI into cloud LLM apps for offline support, better privacy, and lower API costs.
local-ai-use Route image generation, text-to-speech, and speech-to-text through a local AI Server to reduce token/cost usage.

Hardware-native skills

Diagnose, configure, and tune AMD devices directly.

Skill What it does
apu-memory-tuner Inspect and tune the shared-vs-dedicated memory split (GTT / UMA Frame Buffer) on AMD Ryzen APUs.
rocm-doctor Detect driver / kernel / ROCm / framework mismatches and propose fixes.
mi-tuner Opinionated inference tuning for MI accelerators, including TunableOp, FSDP, and FlashAttention.
gfx-target-chooser Pick the right gfx942 / gfx90a / gfx1100 target and matching compiler flags.

Cross-stack porting

Bring existing workloads onto AMD.

Skill What it does
cuda-to-hip Port CUDA kernels with hipify and flag anything that needs manual review.
vllm-rocm Stand up vLLM on AMD with the right environment variables and model configurations.
pytorch-rocm-setup Get a known-good PyTorch + ROCm stack running on a target node, end to end.

Profiling and delivery

Close the loop from trace to fix to ship.

Skill What it does
rocprof-capture Capture and interpret a rocprof trace for a workload.
omniperf-tune Run omniperf, locate the bottleneck, and suggest the fix.
quark-quantize Quantize PyTorch and ONNX models with AMD Quark (INT4/INT8/FP8/MX), pick the right scheme and calibration, and export for AMD deployment.

Skills land incrementally; see Status for what is available today.

A federated catalog

The AMD stack is large and moves fast. ROCm, HIP, Ryzen AI, and framework integrations each have their own team, release cadence, and validation matrix. A single monorepo of skills, maintained by one central team, would always be a step behind.

So skills here are federated: each skill is owned and versioned by the team that owns the product it describes, and this repository is the catalog that brings them together.

                ┌─────────────────────────────────────────────────────┐
                │                amd/skills (this repo)               │
                │                                                     │
                │   skills/         catalog/         .*-plugin/       │
                │   in-repo skills  pointers         agent manifests  │
                └──────────────────────┬──────────────────────────────┘
                                       │  one install
                                       ▼
                              your AI coding agent
                                       ▲
                                       │  resolves pointers to
       ┌───────────────┬───────────────┼───────────────┬────────────────┐
       │               │               │               │                │
   ROCm/ROCm       ROCm/HIP        Ryzen AI repo   lemonade-sdk    ...more
  rocm-doctor/    cuda-to-hip/    ryzen-ai-tools/   local-ai-app-   product
   gfx-target-...  triton-amd-...  ...               integration/    repos

Each skill stays close to the engineers who ship the underlying product, the CI that validates it, and the release tag that pins it.

This repo also acts as an incubator: a skill can start its life under skills/ here to iterate quickly, then graduate to its product repo and be re-pointed from catalog/ once it has a clear owner, with no change for installed users.

What this means for you

  • One install, full coverage. You add this repository through the plugin flow of your agent and you get the whole AMD catalog, so you do not need to track and install skills product by product.
  • Skills update with the products they describe. When ROCm cuts a new release, the ROCm team updates the ROCm skills as part of that release. You see the new behavior the next time you pull the catalog.
  • Skills you can trust. Each skill is signed off by the team that owns the underlying product, not assembled second-hand by a separate documentation team.

What this means if you contribute

  • In-repo skills (Path A) are best for cross-cutting workflows that do not have a natural product home.
  • Product-repo skills (Path B) are best for skills that should live and version with a specific product. You add the skill folder to your product repo and open a small PR here that registers it in catalog/ with a pinned tag. CI validates the linked skill against the same rules as in-repo skills, and the central plugin manifests surface it through the same one install.

See CONTRIBUTING.md for the step-by-step contribution flow for each path.

Repository layout

skills/             # Skills authored in this repository
catalog/            # Manifest pointers to skills that live in product repositories
.cursor-plugin/     # Cursor plugin manifest
.claude-plugin/     # Claude Code marketplace manifest
.github/workflows/  # CI for validating skills and manifests
scripts/            # Tooling for publishing and regenerating manifests

Manual Installation

AMD Skills are compatible with Cursor, Claude Code, OpenAI Codex, and Gemini CLI. The general flow:

Cursor

Install the AMD plugin from this repository through the Cursor plugin flow. The repo ships a .cursor-plugin/plugin.json so skills are discoverable as soon as the plugin is enabled.

Claude Code

Register this repository as a plugin marketplace, then install individual skills:

/plugin marketplace add amd/skills
/plugin install <skill-name>@amd/skills

OpenAI Codex

Copy or symlink the desired folders from skills/ into one of Codex's standard skill locations (for example $REPO_ROOT/.agents/skills or $HOME/.agents/skills). Codex will discover the SKILL.md files automatically.

Gemini CLI

A gemini-extension.json will be provided so the repo can be installed as a Gemini CLI extension:

gemini extensions install https://github.com/amd/skills.git --consent

Using a skill

Once a skill is installed, reference it in plain language while talking to your agent. For example:

  • "Use AMD Skills to integrate local AI capabilities into my app with Embeddable Lemonade."
  • "Use AMD Skills to get PyTorch running on this MI300X node."
  • "Use AMD Skills to convert these CUDA kernels and flag anything that needs manual review."

The agent loads the matching SKILL.md and any helper scripts, then carries out the task. In most cases the agent will pick the right skill on its own from the description; explicit invocation is a fallback, not a requirement.

Contributing a skill

We welcome contributions from AMD engineers and selected partners. There are two paths, matching how the catalog is organized:

  • Path A: In-repo skills. Authored directly under skills/ in this repository. Best for cross-cutting workflows that do not have a natural product home.
  • Path B: Product-repo skills. Authored in a product repository and registered here through catalog/ with a pinned tag. Best for skills that should ship and version with a specific product (HIP, ROCm, Ryzen AI, Lemonade, etc.).

See CONTRIBUTING.md for step-by-step instructions, the full authoring guide, and the rules CI enforces on every pull request.

Status

This repository is in its early days. In-repo skills include skills/local-ai-app-integration/ and skills/local-ai-use/, seeding the Application integration focus area, and skills/apu-memory-tuner/ and skills/rocm-doctor/, seeding the Hardware-native focus area. The remaining skills are being built out incrementally alongside manifests and CI. Expect rapid iteration.

License

Released under the MIT License. See LICENSE for details.