Scientific data are increasingly consumed by AI agents, yet current dataset representations are designed for human users and offer limited support for autonomous discovery, interpretation, and invocation. This gap stems from data being scattered across heterogeneous repositories and from the lack of structured dataset descriptions. To bridge this gap we introduce the Scientific Data Skill (SciDSK), an agent‑ready representation that packages dataset descriptions, scientific context, file organization, usage procedures, quality checks, and provenance while keeping the original data in its repository. We define a structured SciDSK specification and build a systematic construction pipeline that grounds each SciDSK in authoritative dataset records and supporting materials. In addition, we launch the Scientific Data Skill Bank, a unified platform that publishes SciDSK resources across six scientific disciplines, offering package access, persistent identifiers, and traceability to source datasets. Evaluation on a retrieval benchmark for dataset discovery and controlled cases for dataset interpretation shows that SciDSK improves agent‑driven discovery and provides more precise, actionable support for interpretation, confirming the value of organizing dataset‑specific knowledge in an agent‑ready form.
Blogger's Review: SciDSK provides a much‑needed standard for linking AI agents with scientific data. Its modular design and the Skill Bank make it practical for cross‑disciplinary data sharing and automated analysis. Open‑sourcing the construction tools and expanding coverage could further accelerate intelligent use of research data.