Google Ads Master

• You are setting up a new Google Ads account or onboarding an existing account and need to collect configuration, conversion actions, budgets, and guardrails before touching anything.
• You want campaign-level or keyword-level performance reports—CTR, ROAS, Search Impression Share, CPA—interpreted against your actual primary conversion action rather than mixed UI totals.
• You need to investigate low CTR, diagnose conversion-value discrepancies, or analyze search term and geography data to make bid/negative/split decisions.
• You want to create new Search campaigns, ad groups, or keywords in a paused state from a structured spec, with a post-write GAQL readback to confirm what was created.

• You need to manage non-Search campaign types or platforms outside of Google Ads—this skill is scoped to Google Ads Search workflows.
• You want fully automated, hands-off campaign optimization with no human confirmation step; this skill requires explicit permission before any write operation.
• You have not yet completed minimum onboarding (customer ID, auth, reporting timezone, primary conversion action)—the skill will prompt for these before proceeding with analysis or writes.

Before any analysis or account mutation, the skill collects a minimum required dataset: customer ID, OAuth/auth path, developer token, reporting timezone, business model, primary conversion action, ROAS/CPA targets, and write permissions. This prevents acting on incomplete or mismatched account context.

All campaign, ad group, and keyword creation defaults to PAUSED status unless the user explicitly requests otherwise. Budgets and bids are never modified without confirmation, and every write is followed by a GAQL readback to summarize IDs and statuses.

Performance analysis is anchored to the user's confirmed primary conversion action—never assumed to be Purchase, lead, or signup. Mixed Google Ads UI conversion totals are flagged as unreliable until the user confirms which actions count.

Four scripts ship with the skill: a shared client/config helper, a campaign summary report (CTR/ROAS/IS diagnostics), a keyword-level primary-conversion report, and a paused Search campaign builder from a JSON spec. All scripts use the google-ads Python gRPC client.

Recommendations follow a consistent workflow: segment by intent bucket using the user's actual keywords and market, apply keep/observe/reduce/pause/split/add-negatives actions, and always reference campaign and keyword names alongside metrics for actionability.

The skill references internal GAQL patterns and API caveats to avoid known REST incompleteness issues, preferring the gRPC client. Post-write readbacks confirm campaign id/name/status/budget, ad group id/status, and keyword text/match/status.

A new user connects their Google Ads account and the skill walks through the onboarding checklist—collecting customer ID, auth, conversion action, targets, and guardrails—before running report_campaigns.py to produce a baseline CTR, ROAS, and Impression Share summary.

An advertiser suspects their primary conversion action is diluted by micro-conversions in the UI. The skill runs purchase_keywords.py scoped to the confirmed primary action and surfaces which keywords are driving real outcomes versus inflated counts.

Campaign CTR has dropped and the cause is unclear. The skill pulls search term and keyword data, diagnoses whether the issue is ad relevance, match type drift, or Rank Lost Impression Share, and recommends specific keep/pause/split/negative actions with metric evidence.

A user provides a campaign brief with budgets, bidding strategy, keyword list, match types, negatives, ad copy, and URLs. The skill formats a JSON spec, runs create_search_skeleton.py in dry-run first, then creates the structure PAUSED and reads back all IDs and statuses for confirmation.