SKILLFOUNDRY: Building Self-Evolving Agent Skill Libraries from Heterogeneous Scientific Resources

2026-04-05

AI Summary

Scientific knowledge fragmentation hinders agents from effectively leveraging dispersed procedural knowledge. To address this, this work proposes SkillFoundry, a novel framework that, for the first time, automatically extracts, compiles, and validates executable skill packages from heterogeneous scientific resources to construct a self-evolving agent skill repository. Organized around a domain knowledge tree, SkillFoundry integrates resource mining, operational contract extraction, and closed-loop validation mechanisms. Notably, 71.1% of the generated skills are entirely new, substantially expanding coverage beyond manually curated skill sets. The framework enhances coding agent performance on five out of six MoSciBench datasets and demonstrates significant improvements in cell type annotation and scDRS genomic tasks.

Abstract

Modern scientific ecosystems are rich in procedural knowledge across repositories, APIs, scripts, notebooks, documentation, databases, and papers, yet much of this knowledge remains fragmented across heterogeneous artifacts that agents cannot readily operationalize. This gap between abundant scientific know-how and usable agent capabilities is a key bottleneck for building effective scientific agents. We present SkillFoundry, a self-evolving framework that converts such resources into validated agent skills, reusable packages that encode task scope, inputs and outputs, execution steps, environment assumptions, provenance, and tests. SkillFoundry organizes a target domain as a domain knowledge tree, mines resources from high-value branches, extracts operational contracts, compiles them into executable skill packages, and then iteratively expands, repairs, merges, or prunes the resulting library through a closed-loop validation process. SkillFoundry produces a substantially novel and internally valid skill library, with 71.1\% of mined skills differing from existing skill libraries such as SkillHub and SkillSMP. We demonstrate that these mined skills improve coding agent performance on five of the six MoSciBench datasets. We further show that SkillFoundry can design new task-specific skills on demand for concrete scientific objectives, and that the resulting skills substantially improve performance on two challenging genomics tasks: cell type annotation and the scDRS workflow. Together, these results show that automatically mined skills improve agent performance on benchmarks and domain-specific tasks, expand coverage beyond hand-crafted skill libraries, and provide a practical foundation for more capable scientific agents.

Problem

• scientific agents
• procedural knowledge
• heterogeneous resources
• skill libraries
• knowledge fragmentation

Innovation

• self-evolving agents
• skill mining
• heterogeneous scientific resources
• executable skill packages
• closed-loop validation