AI-Research-SKILLs
Comprehensive open-source library of AI research and engineering skills for any AI model. Package the skills and your claude code/codex/gemini agent will be an AI research agent with full horsepower. Maintained by Orchestra Research.
A collection of 98 plug-in skill modules that give AI coding agents like Claude Code or Cursor deep expertise in machine learning research workflows, from fine-tuning models to writing academic papers.
This repository is a collection of 98 add-on modules, called skills, that you can install into AI coding agents like Claude Code, Cursor, or Gemini CLI to give them deep expertise in machine learning research tools and workflows. The goal is to let an AI agent handle the full lifecycle of an AI research project: finding ideas, running experiments, and writing a paper, without the human having to guide each step.
The skills are organized into 23 categories covering the main areas of modern AI research work. Categories include fine-tuning language models, training models across multiple machines, making models faster for deployment, building retrieval systems, working with multimodal data like images and audio, ensuring model safety, and writing academic papers with LaTeX. Each skill is documented with real code examples and practical guides for a specific framework, such as vLLM for fast inference, DeepSpeed for distributed training, or TRL for reinforcement learning from human feedback.
A special skill called autoresearch sits at the top and acts as a coordinator. When you give an agent a research question, autoresearch figures out which sub-skills to call and in what order, routing through literature review, experimentation, and paper writing as needed.
Installation is done via a one-line command using npx, or through the Claude Code plugin marketplace if you use that tool. The installer detects which AI agents you have installed and puts the skills where each agent can find them. You can install all 98 at once, pick a category bundle, or install individual skills.
The project is maintained by Orchestra Research and is licensed under MIT. The full README is longer than what was shown.
Where it fits
- Install all 98 ML research skills into Claude Code or Cursor in one command and give your AI agent expert knowledge of tools like vLLM, DeepSpeed, and TRL.
- Run a full AI research pipeline, literature review, experiments, and paper writing, by prompting the autoresearch coordinator skill with a single research question.
- Install only the fine-tuning or distributed training skill bundle to give your agent targeted expertise without all 98 modules.
- Have an AI agent write a complete LaTeX research paper draft by using the paper-writing skill with your experimental results.