Ontology-Driven Agents Skills for Selection of Building Datasets with various specifications

Abstract. Selecting the correct dataset among Taiwan's nine building specifications requires knowledge of production lineage, legal basis, completeness constraints, and spatial-unit semantics that are rarely represented in machine-readable form, leaving large language model (LLM) agents prone to structurally unsuitable recommendations. This study proposes a framework that integrates OWL ontology reasoning with an agent skills architecture to address this gap. A four-dimensional attribute framework first characterizes the nine datasets across data identification, spatiotemporal characteristics, production and legal basis, and usability, separating properties suitable for formal reasoning from descriptive reference content. These formalizable properties are encoded in an OWL 2 DL ontology, where class hierarchies including a multiple-inheritance HybridData class allow a HermiT reasoner to automatically derive suitableFor and notSuitableFor relationships through subsumption, without manually authored rules. The inferred relationships are compiled into an agent skills decision tree (SKILL.md) with on-demand reference documents, letting an LLM agent return ontology-grounded recommendations for natural language queries. Evaluated against an unsupported baseline LLM on 37 direct, ambiguous, and professional queries, the ontology-grounded agent scored a perfect 10.0/10 average and won or tied every comparison, while the baseline averaged 7.5/10 and, in the sharpest case, recommended a dataset the ontology explicitly marks unsuitable. Two demonstration cases further show the framework handling both single-domain suitability rejection and cross-domain, constraint-aware integration.