菜单
Microsoft Research's BioEmu — generates protein conformational ensembles from amino acid sequence on a local GPU. Use when users mention BioEmu by name, want to predict protein dynamics or flexibility from sequence, or are setting up / running / troubleshooting BioEmu locally.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/microsoft/vibe-kit --skill msresearch-bioemu命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Aurora — Microsoft's foundation model for the Earth system. Use this skill for both inference (regional weather/climate forecasting, Norway example, ERA5/CDS data, wind/temperature/pressure mapping) and fine-tuning Aurora on new variables, regions, or domains (training, evaluation, checkpoint extension, gradient/loss debugging). Triggers on Aurora, Microsoft Aurora, foundation models for Earth systems, ERA5/CDS/NetCDF weather data, weather forecast validation, and adapting Aurora to new use cases.
Microsoft Research's Dayhoff — generates novel protein sequences and scores variants using hybrid Mamba+Transformer models trained on the 3.34B-sequence Dayhoff Atlas. Use when users mention Dayhoff by name, want to generate protein sequences from scratch or from a prompt, score mutations / variants / fitness without experimental data, design sequences in a protein family from homologs, scaffold sequences around a motif, or are setting up / running / troubleshooting the Dayhoff prototype locally.
MatterGen workflows for property-conditioned inorganic crystal generation, MatterSim triage, adapter fine-tuning, and Azure AI Foundry hosted deployment. Use when users ask about MatterGen, materials discovery, crystal generation, property-conditioned diffusion (bulk modulus, band gap, magnetic density, HHI), MatterSim evaluation, or Azure AI Foundry materials deployment.
基于 SOC 职业分类
正在显示 SKILL.md
| name | msresearch-bioemu |
| description | Microsoft Research's BioEmu — generates protein conformational ensembles from amino acid sequence on a local GPU. Use when users mention BioEmu by name, want to predict protein dynamics or flexibility from sequence, or are setting up / running / troubleshooting BioEmu locally. |
| license | MIT |
pip install bioemu[cuda] and python -m bioemu.sample directly from the upstream package.assets/reference-app/. Run its bundled score/ Flask server locally on :5001, point the proxy backend at it via .env, and use the React frontend for visualization.~/.cache/colabfold/ (CLI) or /app/colabfold_cache (Docker).nvidia-container-toolkit), or system Python for a bare-metal score/ run; Node.js 18+ for the frontend.docs/about-bioemu.md when users ask what BioEmu is, how it works, or need scientific background.docs/quick-start.md to get a first ensemble locally — Path A (CLI smoke test) before Path B (reference app).docs/application-patterns.md.docs/troubleshooting.md.nvidia-smi and confirm a CUDA-capable GPU is visible. If none is present, warn the user that BioEmu is unusable beyond ~10-residue toy sequences and stop — do not proceed to install.pip install bioemu[cuda], pip install -r requirements.txt, and npm install --legacy-peer-deps are all slow on first run. Do not interrupt installs.score/ server, proxy backend, and frontend each run in their own terminal and must stay running. Never run other commands in a terminal hosting a live server..env keeps its AZURE_BIOEMU_* variable names because the proxy reads those exact names — but AZURE_BIOEMU_ENDPOINT should point at http://localhost:5001/score and AZURE_BIOEMU_KEY can be any non-empty string (the local score/ server doesn't enforce auth). Explain this when guiding users; never ask them to paste secrets into chat.docs/about-bioemu.md), end your response with a concrete offer to advance the user along the Learning Path. Default phrasing: "Want to try BioEmu? I can walk you through running it locally — three commands gets you a first ensemble on your GPU, and from there we can wire up the full reference app UI if you want it." Adapt wording to context, but never end an explainer response without a concrete next-step offer.| Doc | When to load |
|---|---|
docs/about-bioemu.md | User asks what BioEmu is, how it works, performance metrics, limitations, or the scientific FAQ |
docs/quick-start.md | User wants to run BioEmu locally (CLI or full reference app) |
docs/application-patterns.md | User wants code examples for sampling, MDTraj analysis, output file formats, or AlphaFold comparison |
docs/troubleshooting.md | User hits an error, missing GPU, weight download stall, port conflict, or MSA timeout |
docs/about-bioemu.md — Understand what BioEmu is and why it mattersdocs/quick-start.md Path A — Three-command CLI smoke test on your GPUdocs/quick-start.md Path B — Run the full reference app against a local score/ serverdocs/application-patterns.md — Build your own sampling and analysis workflowsassets/reference-app/ — Self-contained React + Flask + score/ Docker app. This is Path B; no separate clone needed.