What problem does it solve? Research papers often cite external statistics (unemployment rates, GDP series, market data) that are typed in by hand without a verifiable source, making claims impossible to audit or reproduce. This Skill enforces a provenance-first workflow where every external data lookup becomes a stored, dated, versioned artifact linked to the paper claim it supports. ## Core Features & Use Cases - Provenance-First Retrieval: Records the source database, dataset ID, endpoint, retrieval date, version/release, units, and coverage period for every external series. - Use Classification: Classifies each lookup as contextual, inferential, or illustrative, with inferential use triggering the full provenance path including schema validation and vintage checks. - Scripted Transformations Only: Prohibits hand-edited spreadsheets between source and artifact, requiring all transformations to be scripted. - Use Case: A World Bank series is about to enter a panel model as a covariate. The Skill classifies it as inferential, stores the raw response with retrieval metadata, and checks that the series vintage matches the panel window before the value enters the model. ## Quick Start Retrieve the UK unemployment rate series from ONS for my paper's background section and store it as a dated provenance artifact.
