agentic-science-worker
A framework for evaluating core capabilities and benchmarking autonomous agents in computational materials science.
Install & Use
Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:
Help me install this AI Skill: agentic-science-worker. It is used for: A framework for evaluating core capabilities and benchmarking autonomous agents in computational materials science. Full Skill content: https://321skill.com/skills/agentic-science-worker/raw/index.md Read that page and install it.
The prompt includes a link to the full Skill content. You can also view the full content.
Agentic Science Worker (ASW) has been renamed and refactored into Caliber. This project addresses the lack of a standardized, evaluable framework and test benchmark for autonomous agent capabilities in computational materials science. It provides a core competency library encompassing 17 essential scientific skills and agent collaboration specifications, paired with a benchmark suite for assessing the performance of autonomous materials science agents. This suite features a three-dimensional scoring system, native model adaptation, and versioned task sets.
Users can integrate its core competency library to enhance an AI agent's specialized reasoning and computational abilities in materials science, or utilize its benchmark suite to evaluate and compare the performance of different agent models on materials science tasks. The project offers standardized tools and methodologies for building and evaluating AI agents in specialized domains.
It is well-suited for academic researchers in computational materials science, developers building specialized scientific agents, and data analysts requiring algorithm capability assessment and comparison. These roles can leverage the project to accelerate research, develop more reliable AI tools, or conduct objective performance evaluations.
Please note: The project has been fully migrated to the new Caliber repository; the current repository is archived. All new users are advised to visit the new repository for the latest version and ongoing support. Historical versions, releases, and the early benchmark suite from the original repository remain accessible for reference.
Key Features
Unlike general-purpose scientific agents or code generation tools, it is specifically tailored for computational materials science. It offers a structured set of 17 core skills with accompanying agent collaboration specifications (AGENTS.md) and integrates a dedicated benchmark framework for this domain, featuring three-dimensional scoring and versioned tasks.
Limitations
Not suitable for scenarios requiring general-purpose programming, research outside the materials science domain, or simple code generation.
FAQ
Is this repository still usable?
This repository is archived. The project has been migrated and renamed to Caliber. Please visit the new repository for the latest version.
What does Caliber primarily include?
It includes a core competency library (17 skills) for computational materials science agents and a benchmark suite for evaluating such agents.
Installation guide for AI assistants
If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:
Visit https://321skill.com/skills/agentic-science-worker/raw/index.md to read the original Skill definition (Markdown format) for agentic-science-worker, and install it according to the instructions.
Raw Markdown URL for AI: /skills/agentic-science-worker/raw/index.md