What problem does it solve? Answering "how are my ads doing" safely is hard: an AI assistant near an ad account can either invent metrics from memory or apply an unapproved "fix" that spends real money. This Skill turns that ask into a disciplined read-only review where every number names its campaign, window, and source, and no change is ever applied without an explicit yes. ## Core Features & Use Cases - Read-only campaign review: Pulls spend, clicks, conversions, and anomalies through the configured Google Ads read surface, never mutating bids, budgets, targeting, or creative. - Two-slice comparison: Requires two campaigns, or one campaign across two windows, so movement is demonstrated rather than asserted. - Anomaly flagging with evidence: Flags sudden spend shifts, zero-conversion spend, and disapproved or limited statuses, each tied to a campaign, metric, and window. - Use Case: A founder asks their Bot "how did my ads do last month?" The Skill pulls two 30-day windows per campaign, flags a zero-conversion spend spike with its campaign id, ranks candidate fixes by likely impact, and ends with "No changes made." ## Quick Start Ask the assistant to review how your Google Ads campaigns performed over the last 30 days and flag anything unusual without changing anything.
