ai4science-studio: Skill for Cursor

Discovering models in AI4Science Studio

Use this skill when a user asks what models are available, wants to compare models, or asks questions like "what can I do with this repo?"

Quick discovery

Read models.yaml at the repo root (generated from per-model model.yaml via make fix). It lists every model with:

• slug — folder name
• domain — earth_science, material_science, healthcare, physics_simulation, protein_folding
• hf_id — Hugging Face model id (or N/A)
• license — SPDX id
• summary — short one-liner for tables
• task — fuller description from the manifest
• tasks_available — list of recipe tasks (inference, train, finetune, ensemble)
• path — relative path to model folder

Answering "what models are available?"

Parse models.yaml and present a table:

Model	Domain	Task	Available recipes
StormCast	Earth science	Weather prediction	inference, ensemble
...	...	...	...

Use the summary field for the Task column (not the longer task field).

Filtering

When the user asks for a subset:

• By domain: filter on the domain field
• By task type: filter on tasks_available (e.g. "which models support fine-tuning?" → filter for finetune in tasks_available)
• By license: filter on license (e.g. "which models are MIT licensed?")
• By HF availability: filter on hf_id != N/A for models with HF weights

Deep-dive on a model

When the user wants details on a specific model:

1. Read <path>/model.yaml for structured metadata
2. Read <path>/README.md for prose description
3. Read <path>/examples/README.md for available scripts
4. List <path>/recipes/ subdirectories for available recipes

Comparing models

When the user asks to compare two or more models:

1. Read each model's model.yaml
2. Present a comparison table with columns: name, domain, HF id, license, tasks, container image, validated hardware, VRAM

Checking readiness