skills vs AREX-Skill
Agent-ready workflows for AMD and machine learning
skills provides task-focused guidance for coding agents working with AMD hardware and software, while AREX-Skill packages workflows drawn from machine-learning repositories and research. The main difference is scope: skills focuses on AMD-specific tasks, whereas AREX-Skill supports broader ML engineering and research workflows, with tools to manage and route its skills.
skills: Give AI Agents AMD Workflow Guidance
AMD Skills is a catalog of task-focused instructions and tools that help coding agents work with AMD hardware and software. Install selected skills into compatible agents when you need guidance for workflows such as local AI, ROCm troubleshooting, or LLM serving.
AREX-Skill: Give Coding Agents Reusable ML Workflows
AREX-Skill is a library of executable, agent-readable workflows distilled from machine-learning repositories and research. Use it to help coding agents run, validate, and recover ML tasks with less unguided trial and error.
| skills | AREX-Skill | |
|---|---|---|
| Language | Python | Python |
| License | MIT | Apache-2.0 |
| Stars | 386 | 327 |
| Forks | 39 | 25 |
| Last analyzed | Oct 3, 2026 | Oct 4, 2026 |
Key differences
- skills covers AMD workflows across Ryzen AI, ROCm, AMD Instinct GPUs, and EPYC CPUs; AREX-Skill draws on machine-learning repositories and research.
- skills supports installation through the skills CLI and targets Cursor, Claude Code, OpenAI Codex, and Gemini CLI; AREX-Skill uses DisCo and lists Codex, Claude Code, and Pi.
- AREX-Skill routes guidance through skill graphs and includes workflow validation and recovery steps; skills provides focused instructions and, where needed, scripts or references.
- skills is licensed under MIT; AREX-Skill's repository license is Apache-2.0, and individual AREX-Skill skills may have separate licenses.
- skills notes that some catalog entries may be planned or in tech preview; AREX-Skill says its full technical report for benchmark results is forthcoming.
- Both projects report Python as their language, and neither is archived.
Choose skills if you…
- need agent guidance for AMD software or hardware tasks such as ROCm troubleshooting or LLM serving.
- want to install selected AMD-focused skills into a compatible coding agent.
- work with AMD products and can check each skill's scope and product status.
Choose AREX-Skill if you…
- want agent workflows distilled from machine-learning repositories or research.
- need skills with validation steps and guidance for recovering from failed experiments.
- plan to create, verify, refresh, or export skills using DisCo and can configure a model provider.
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.