What is science-skills?

Science Skills is a collection of add-on capabilities that give AI agents specialized knowledge for scientific research. Think of it as a toolbox an AI can use to help with tasks in genomics, structural biology, chemistry, and literature search, drawing on data from over 30 established databases and tools. Each skill is a package of instructions, helper scripts, and reference material. When an AI agent equipped with these skills receives a scientific question, it reads the skill's instructions and uses the provided scripts to query databases like UniProt or ClinVar. The design aims to make agents more accurate by grounding their responses in real scientific data, and more efficient by reducing the amount of processing power needed to arrive at an answer. The primary audience is researchers, scientists, and students who use AI tools to accelerate their workflows. For example, a genetics researcher could use a skill connected to AlphaGenome to analyze genomic data, or a pharmacologist could use chemistry-focused skills to look up compound information. The skills integrate directly into Google Antigravity, an AI application, where users can enable the Science plugin during setup or through settings. Some skills require API keys for certain databases, and the agent guides users through obtaining them. The project is built in Python and uses a package manager called uv to handle dependencies automatically the first time a skill runs. For users who want to tailor skills to their specific needs, the repository supports customization. Rather than editing the installed plugin directly (which would get overwritten on updates), users can place modified or entirely new skills in a personal directory on their computer. This makes the system extensible for labs or individuals with niche requirements beyond the bundled offerings.