What problem does it solve? Manually verifying a full BigQuery agentic pipeline—data generation, conversational agents, verified golden queries, and Gemini Enterprise A2A bindings—is slow and error-prone, and leftover cloud resources create cost and hygiene risks. This Skill automates the entire validation cycle against an isolated local server and tracks every provisioned asset for cleanup. ## Core Features & Use Cases - Isolated E2E Execution: Spawns a temporary Node.js backend on an auto-allocated local port and runs the full five-stage pipeline (research, lakehouse generation, Bronze deployment, Conversational Agent creation, Gemini Enterprise A2A registration). - Verified Golden Queries: Generates and compiler dry-run validates three verified queries (ranking, time series, distribution) persisted to the BigQuery Data Agent's published and staging contexts. - Manifest Ledger & 1-Click Cleanup: Logs all created datasets, tables, and agents to a timestamped JSON manifest, with dry-run, single-run, or all-runs teardown of GCP resources. - Use Case: A data engineer validating a demo for a Brazilian telecom runs the runner with the company domain and project IDs, receives console links to the live Conversational Agent and Gemini Enterprise assistant, then deletes all provisioned assets with one cleanup command. ## Quick Start Run the isolated end-to-end test for company www.vivo.com.br against my BigQuery data project and Gemini Enterprise project, then clean up all provisioned resources afterward.
