What problem does it solve? Research teams cannot find the datasets they need because catalogs either duplicate stale metadata centrally or mirror pipelines and org charts, making search fail. This Skill guides the design of a dataset catalog or marketplace where researchers and agents go from a need statement to a pinned, accessible dataset version without leaking unauthorized metadata. ## Core Features & Use Cases - Metamodel and domain design: Define domains from knowledge areas rather than pipelines or org charts, with vertical, horizontal (lineage-to-checkpoint), and relational browse dimensions. - Catalog vs marketplace architecture decision: Choose push-based catalog, pull-based marketplace, or hybrid using a decision table covering self-registration, freshness, and incentives. - Governance and search specification: Produce role models, three-tier glossary governance (folksonomy, taxonomy, thesaurus), recall/precision benchmarks, and identity-aware search that filters before retrieval. - Use Case: A research division wants agents to discover training data via MCP-style APIs. Use this Skill to specify the agent-consumable surface: permanent URIs, structured metadata schemas, policy-partitioned vectorized summaries, and certification status inline in results. ## Quick Start Use the dataset-catalog-designer skill to design a dataset catalog for our research division, including the domain map, marketplace architecture decision, role assignments, and a benchmark query set for measuring search recall and precision.
