From 46 Heterogeneous Star Catalogs to One Unified Age Database: An AI-Assisted Curation Workflow

We present a case study of AI-assisted curation of 46 published asteroseismic age catalogs into a unified database of 3.85 million stellar records (3.38 million with age estimates). Using Claude Code with the ARIS skill ecosystem, the workflow automated literature discovery across VizieR TAP and ADS, multi-source download orchestration (six repository types), column standardization across $\sim$200 unique identifiers, and method classification via automated reading of 22 arXiv Methods sections. Key failure modes---hallucinated column mappings, misclassified age derivation methods, undetected bibcode mismatches---were caught by a four-layer verification protocol. The AI-assisted workflow reduced the curation timeline from 8--12 weeks to approximately 2 weeks, while correcting 12 method misclassifications that would have propagated systematic errors into downstream analyses.