What problem does it solve? Single-pass text-to-SQL generation often produces inconsistent or incorrect queries when questions are ambiguous or the data model context is incomplete. This Skill runs every data question through a multi-agent consensus protocol so the final SQL is cross-validated before execution. ## Core Features & Use Cases - Five-Agent Quorum: Spawns 5 sub-agents with identical context blocks and compares their SQL outputs for semantic equivalence across tables, joins, aggregations, filters, and grouping. - Disagreement Diagnosis: Instead of voting, it identifies why agents disagree — ambiguous questions, missing context, conflicting vault definitions, or grain confusion — and resolves the root cause. - Iterative Refinement: Runs up to 3 full quorum rounds with refined context or user clarification, and reports unresolved disagreements instead of guessing. - Use Case: A user asks "What was revenue last quarter?" and agents split between gross and net revenue interpretations. The Skill detects the ambiguity, asks the user to clarify, and regenerates a consensus query. ## Quick Start Ask a data question and have the agent run the SQL consensus protocol to generate a validated query from the vault context.
