Microsoft Research Agent Skills — specialized knowledge, workflows, and demos that teach AI coding assistants how to work with Microsoft Research projects.
This repository hosts Microsoft Research Agent Skills — drop-in skills for AI assistants like GitHub Copilot and Claude that teach them how to work with Microsoft Research projects. Each skill includes documentation, working demos, and reference implementations.
| Skill | Research Area | Description |
|---|---|---|
| Aurora | Weather Forecasting | Foundation model for Earth systems — predicts weather, air quality, and ocean waves |
| MatterGen | Materials Science | Property-conditioned crystal generation for materials discovery |
| BioEmu | Computational Biology | Protein conformational ensemble generation and dynamics analysis |
| Dayhoff | Computational Biology | Protein sequence generation and mutation scoring using hybrid Mamba-Transformer models |
| Promptions | AI/ML Tools | Dynamic control generation and structured prompt middleware for AI copilots |
Each skill has its own directory with:
- SKILL.md — What the skill does and when it activates
- README.md — Overview and links to documentation
- docs/ — Detailed guides, tutorials, and reference material
- assets/ — Working demos, sample data, and scripts
Start here: Pick a skill from the table above and open its directory.
To install a skill for use with your AI assistant:
npx skills add microsoft/vibe-kit/skills/name-of-skillReplace name-of-skill with the skill directory you want to install.
These skills are part of Microsoft Research Agent Skills — designed to be loaded by AI coding assistants. When you ask your assistant about a topic (e.g., "help me run an Aurora weather forecast"), it loads the relevant skill to get specialized knowledge and workflows.
Skills are defined in SKILL.md files with activation conditions that tell the assistant when to use them.