SciVisAgentSkills: Design and Evaluation of Agent Skills for Scientific Data Analysis and Visualization

2026-06-03

AI Summary

Existing general-purpose coding agents struggle to reliably execute multi-step scientific visualization (SciVis) workflows due to a lack of domain-specific tool knowledge. This work presents the first systematic construction and evaluation of a structured skill library tailored for SciVis, enhancing agent capabilities on platforms such as ParaView, napari, VMD, and TTK by encoding environmental assumptions, tool usage patterns, and domain-specific heuristics. Agents powered by Codex and Claude Code demonstrate significantly improved task success rates on the newly introduced SciVisAgentBench benchmark, which comprises 108 expert-designed tasks. Furthermore, these agents exhibit higher token efficiency across diverse tool and model configurations, underscoring the critical synergy between procedural knowledge and execution frameworks in enabling robust SciVis automation.

Abstract

Recent advances in agentic visualization have enabled the translation of natural language into executable scientific visualization (SciVis) workflows. While general-purpose coding agents show strong capabilities, they often lack the tool-specific expertise required for SciVis tasks. In this work, we present SciVisAgentSkills, a collection of reusable agent skills that augment coding agents for scientific data analysis and visualization by encoding environment assumptions, tool usage patterns, and domain heuristics across scientific tools such as ParaView, napari, VMD, and TTK. We evaluate these skills on Codex and Claude Code using SciVisAgentBench, a benchmark of 108 expert-designed multi-step tasks. Results show that agent skills improve mean task scores across the evaluated suites, with token-efficiency benefits that depend on the agent harness and tool setting. These findings highlight the importance of structured procedural knowledge for enabling reliable, long-horizon SciVis workflows, while also showing that skills should be studied alongside the execution harness that loads and applies them. The skills are available at https://github.com/KuangshiAi/SciVisAgentSkills.

Problem

• scientific visualization
• agent skills
• tool-specific expertise
• long-horizon workflows
• procedural knowledge

Innovation

• agent skills
• scientific visualization
• procedural knowledge
• tool-specific expertise
• SciVisAgentBench