Benchmarking Security Risk Detection and Verification in Open Agentic Skill Ecosystems | ArxivCSExplorer