What problem does it solve? Large scientific datasets often already live in cloud buckets (gs:// or s3://) and never pass through an upload pipeline, leaving them invisible to catalog search. This Skill registers those datasets by reference in the Chronicle platform and authors the rich, math-bearing metadata layer (PDE, boundary conditions, domain geometry, per-variable shapes and dtypes) that makes them describable, filterable, and searchable. ## Core Features & Use Cases - Register by reference: Create a ready dataset asset from an existing gs:// or s3:// URI in a single MCP call, with no byte upload, presign, or finalize steps. - Author and patch metadata: Write a long LaTeX-bearing description, PDE family and equation, boundary/initial conditions, domain geometry, a per-variable table, and free-form key=value properties; update the mutable metadata layer later without minting a new asset. - List and filter the catalog: Run Postgres-side filtered listing on promoted facets such as n_dims, precision, pde_family, geometry, and size. - Use case: A researcher has a 256^3 Navier-Stokes DNS corpus in a GCS bucket and wants teammates to find it by searching "3D incompressible Navier-Stokes, periodic box, fp64" — this Skill registers the URI and authors exactly that metadata. ## Quick Start Ask the agent to register the dataset already stored at your gs:// bucket URI as a searchable Chronicle dataset with a description of its PDE, boundary conditions, and variables.
