What problem does it solve? Analysts querying Credible workspaces through MCP tools often misuse the object hierarchy, retry expected empty states, or rebuild metrics from raw columns instead of using the modeler's published measures and views. This Skill teaches the correct mental model and retrieval judgment so queries against published Malloy semantic models return numbers that match the team's reporting. ## Core Features & Use Cases - Object model orientation: Maps the full hierarchy from organization and workspace down through environment, package, model, and source with its dimensions, measures, views, and joins. - Retrieval judgment: Covers validating known entity names without a search, treating names as unconfirmed pointers, and when dimensional_value refinement is worth using based on the values_indexed flag. - Empty-state handling: Explains that unpublished workspaces return an expected empty state that must not be retried or reported as a broken model. - Use Case: You ask how many active subscribers existed last month; the Skill guides Claude to get the workspace from list_workspaces, confirm the measure exists via get_context, and prefer the source's own view over hand-written aggregation. ## Quick Start Ask a natural-language question about your Credible workspace data, such as how revenue broke down by region last quarter, and let the published semantic model guide the query.
