What problem does it solve? Database schemas alone cannot carry business semantics like enum meanings, units, NULL semantics, or metric definitions, so AI agents generate wrong SQL. This Skill reads raw artifacts (PDFs, glossaries, handbooks, data dictionaries) and fills those gaps into a Wren project's MDL, cubes, and knowledge files. ## Core Features & Use Cases - Two operating modes: Grill mode interviews the user one question at a time with recommended answers; auto-pilot mode applies inferences directly and escalates only conflicts and high-blast-radius additions like new cubes or relationships. - Ten-category gap catalog: Detects missing enum semantics, units, NULL meanings, magic sentinels, soft-delete filters, synonyms, time conventions, external identifiers, currency rules, and canonical-table preferences, routing each to the correct sink. - Cube proposals: Converts named aggregation metrics (ARR, DAU, churn, NRR) found in raw documents into validated cube YAML with duplication guards against existing cubes and legacy MDL metrics. - Use Case: After generating an MDL, point the Skill at a folder containing your finance handbook and data dictionary; it extracts atomic claims, diffs them against current context, and writes confirmed findings to MDL descriptions, knowledge/rules/, and knowledge/sql/ pairs. ## Quick Start Ask the agent to enrich my Wren project context using the documents in the raw folder, choosing grill mode to review each proposed addition.
