claude-google-ads

Audits and scores an existing Google Ads account across conversion tracking, wasted spend, account structure, keywords/Quality Score, ads & assets (incl. PMax, AI Max, Demand Gen), and settings. Read-only. Generalized fork of the ads-google audit: reads everything from account-context.yaml and applies the operator's guardrails — no brand/vertical hardcoding. Use when the user says "audit", "score my account", "what's wrong with my ads", "wasted spend", "account review", "health check".

Acts on a running account to improve performance: campaign tiering, search-term mining into negatives, budget reallocation, target ROAS/CPA adjustments, bid-strategy changes, asset refresh, and consolidation. Diagnoses root cause (not just symptoms) and produces a dated action plan. Proposes changes — applies via pusher with approval. Generalized fork of a production optimizer; reads account-context.yaml, respects guardrails and margin tiers. Use when the user says "optimize", "improve ROAS", "lower CPA", "add negatives", "raise tROAS", "reallocate budget", "fix underperformers", "weekly ads review".

Pushes a campaign-spec.json (from any builder-*) OR a change-set.json (from optimizer/routine) into Google Ads safely. Reads the JSON contract — never the spreadsheet. Validates, then either exports a Google Ads Editor import file + a guided UI build/change checklist (default, read-only MCP), or performs API mutations when account-context google_ads.api_write is true. Always behind a human approval gate and a spend-cap guard; creates campaigns PAUSED and never enables them. Use when the user says "push", "export to google ads", "upload campaign", "import to editor", "apply changes", "apply the optimizer plan", "go live".

Onboards a business into the claude-google-ads suite by producing account-context.yaml — the backbone every other skill reads. Detects the data source (Shopify/WooCommerce/BigCommerce/CSV via MCP), captures Google Ads account ids and conversion actions, GA4/GSC properties (shared vs per-site), brand and competitor terms, margins, and operator guardrails. Use when the user starts with Google Ads, has no context file yet, or says "set up", "onboard", "connect my store/account", "configure".

Router / orchestrator for the claude-google-ads suite. Use when the user wants to do anything with Google Ads but hasn't named a specific step — "help me with Google Ads", "launch Google Ads for my store", "run google ads", or just /google-ads. Reads account-context.yaml, runs setup if missing, figures out where the account is in the lifecycle, and dispatches to the right sub-skill (setup, measurement, audit, plan, builder-*, pusher, tracker, optimizer, assets, experiments). It orchestrates; it does not do the work.

Validates conversion tracking BEFORE any spend — the #1 failure point in Google Ads accounts. Checks conversion actions (purchase/lead), counting settings, primary vs secondary, double-count risk by source (store channel e.g. Shopify vs GA4 import vs gtag), enhanced conversions, server-side/CAPI, and value tracking — and judges them HOLISTICALLY (which actions each campaign actually fires) before any verdict. Use when the user says "conversion tracking", "is my tracking right", "GA4 import", "enhanced conversions", "why are my conversions wrong", or as the gate before plan/builder.

Plans a Google Ads launch or expansion: how to split budget, which campaign types to run (Search, Branded Search, Performance Max, Demand Gen) and in what order, expected conversions/CPA/ROAS given budget and AOV, and a phased ramp roadmap. Reads context + audit + measurement; forecasts from real account data when available, benchmarks when not. Use when the user says "plan", "budget", "what campaigns should I run", "how should I launch Google Ads", "media plan", "campaign mix", "forecast".

Monitors live campaigns — especially newly launched ones in the learning phase. Reports budget pacing, learning-phase status, delivery, and anomalies (sudden CPC/CPA/ROAS swings, conversion drops, disapproved assets). Observe-only: it alerts, it never changes the account. Reads account-context.yaml. Use when the user says "how's my campaign doing", "pacing", "learning phase", "is it spending", "monitor", "anomaly", "what changed".

Generates Google Ads creative assets shared across campaign types: responsive search ad headlines and descriptions, Performance Max asset-group copy and search themes, Demand Gen image/video briefs, and sitelinks/callouts. Enforces character limits, brand voice from account-context.yaml, and policy/ trademark safety. Use when the user says "ad copy", "headlines", "descriptions", "assets", "creative", "rsa", "asset group copy", "sitelinks", "callouts".

Builds a Branded Search campaign — defensive coverage on the business's own brand terms plus optional competitor conquesting — with trademark-aware ad copy and the brand-exclusion coordination that keeps PMax/Shopping from cannibalizing brand traffic. Emits a human .xlsx plus a campaign-spec.json (branded_search). Reads account-context.yaml. Use when the user says "branded search", "brand campaign", "defend my brand", "competitor conquesting", "bid on my own name".

Builds a Demand Gen campaign blueprint — audience-led, creative-heavy campaigns across YouTube, Discover, and Gmail. Covers audience strategy (your-data, lookalikes, custom segments), image and video creative sets, ad formats, and the upper/mid-funnel measurement caveats. Emits a human .xlsx plus a campaign-spec .json (demand_gen). Reads account-context.yaml. Use when the user says "demand gen", "demand generation", "youtube campaign", "discovery ads", "top of funnel".

Builds a complete Performance Max campaign blueprint: chooses a split strategy (campaign-level + asset- group-level), then designs asset groups, listing groups, ad copy (headlines/long headlines/descriptions), search themes, audience signals, sitelinks, negatives, image/video briefs, budget/bidding ramp, landing- page requirements, and an Editor-ready workbook. Generalized for any business: reads account-context.yaml, no brand/vertical hardcoding. Use when the user says "build pmax", "performance max", "create a pmax campaign", "pmax blueprint".

Builds a standard Search campaign blueprint: ad-group structure (tight theming), keyword lists with match types, responsive search ads (RSAs), negative keyword lists, extensions, and bidding. Emits a human .xlsx plus a campaign-spec.json (ad_groups variant) the pusher can export to Google Ads Editor. Reads account-context.yaml; no hardcoding. Use when the user says "build search", "search campaign", "keyword campaign", "create search ads".

Designs disciplined A/B tests for Google Ads — campaign-level splits (bidding, brand exclusion, new- customer mode, split structure) and asset-group/ad-group-level splits — changing one variable at a time with a feasibility check, clear success criteria, and minimum runtime. Prevents "testing by spawning random campaigns". Reads account-context.yaml. Use when the user says "experiment", "A/B test", "split test", "which works better", "test budget/bidding/audience".

The daily/weekly/monthly/quarterly operating rhythm for a Google Ads account. Orchestrates tracker (observe) and optimizer (act) at the right cadence, remembers when each was last run, auto-surfaces overdue checks, and gates every action behind explicit preconditions (only act when ALL are met, else wait). Use when the user says "daily routine", "weekly review", "what should I check today", "routine", "what's due", "ongoing management", "what can I change now".