菜单
Catch spend spikes, CTR or conversion-rate drops, budget runaway, and metric outliers across campaigns, and set sane alert thresholds. Use when a user asks why spend or performance changed suddenly, wants monitoring or alerts, or something looks off in the numbers.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | anomaly-detector |
| description | Catch spend spikes, CTR or conversion-rate drops, budget runaway, and metric outliers across campaigns, and set sane alert thresholds. Use when a user asks why spend or performance changed suddenly, wants monitoring or alerts, or something looks off in the numbers. |
Find the sudden change before it burns a budget. Reads are free; setting alerts or pausing is a change and waits for approval.
Confirm the account: list_connected_accounts.
Read a stable window and the recent period per platform: pull_<platform>_ads_performance(days=...), or run_gaql_query for Google detail. You need enough history to know what normal looks like before you can call something abnormal.
Rule out the mundane first: a tracking outage reads as a conversion cliff; a new campaign launch reads as a spend spike; a weekend reads as a volume drop. Confirm the cause before recommending action.
set_spend_alert so the next runaway trips an alarm early.Expand targeting without wrecking CPA — lookalike ladders, seed quality, audience layering, and exclusion hygiene across platforms. Use when a user wants to scale beyond exhausted audiences, build lookalikes, fix audience fatigue, or set up exclusion lists.
Build a post-cookie first-party data strategy — Enhanced Conversions, server-side tracking, Consent Mode v2, and activating owned data for targeting. Use when a user wants to recover lost conversion data, set up Enhanced Conversions or CAPI, implement consent mode, or plan for cookie deprecation.
Verify that ad campaigns actually target the geographies they claim — location criteria exist, presence-only matching is set, and click geography matches intent. Use when auditing campaigns, investigating cheap clicks or missing conversions, or reviewing any campaign before and after launch.
Diagnose and improve Google Ads Quality Score — Expected CTR, Ad Relevance, and Landing Page Experience — to cut CPCs and lift Ad Rank. Use when CPCs look high for the vertical, impression share is capped by rank, or keywords sit at QS 5 and below.
Optimize Google Shopping — product feed quality, titles, Merchant Center disapprovals, supplemental feeds, custom-label segmentation, and Shopping vs Performance Max structure. Use when Shopping impression share is low, products are disapproved, or feed titles and labels need work.
Design and run incrementality tests that measure true lift from ads — geo holdouts, platform lift studies, significance math, and iCPA decisions. Use when a user wants to know whether ads actually cause conversions, design a holdout test, or decide budget based on real lift.