Claude Ads: Paid Advertising Audit Skill for Claude Code

A manual audit of a single Google Ads account takes 4-6 hours of senior PPC time. Claude Ads runs the same audit in 10-15 minutes, scores it on a 0-100 weighted scale, and outputs a prioritized action plan across Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, Apple, and Amazon Ads. Built for PPC agencies, in-house marketers, and freelance ad consultants. Local, deterministic, MIT-licensed.

Host support:

Last updated: 2026-05-18 · Version: v1.7.1 · CHANGELOG · Blog: full ad audit breakdown

Two versions of this skill.

• 🌐 Public open-source → AgriciDaniel/claude-ads: MIT, public releases, no membership. Best if you want stable + downloadable.
• 🔒 Community private mirror → AI-Marketing-Hub/claude-ads: early access to upcoming features and direct collaboration with the AI Marketing Hub Pro community. Requires membership.

Who this is for

• PPC agencies running 5+ accounts: audit every account weekly instead of once a quarter. Same time budget.
• In-house marketers owning paid across 4+ platforms: second-pair-of-eyes before exec reviews. No human bias on which platform you favor.
• Freelance PPC consultants: anchor day-1 client scope with a 10-minute audit. Win the engagement before you spend an hour on diagnostic.

What's new in v1.7.0 (Wave 2)

• 3 new sub-skills: /ads amazon (Sponsored Products/Brands/Display, ACOS/TACOS), /ads attribution (AdAttributionKit + GA4 + Consent Mode V2), /ads tracking (sGTM + CAPI Gateway + dedup + hashing).
• 41-test pytest eval harness in tests/: routing snapshots, bidirectional 209-check catalog coverage, scoring math determinism, SSRF regression. Runs in CI on every commit.
• Cross-runtime install matrix: install.sh / install.ps1 --target=<host> with whitelist validation for Claude Code, Codex CLI, Cursor, Windsurf, Gemini CLI, Goose.
• Deep platform rewrites: /ads google for the AI Max era (ai_max_setting.enable_ai_max, AI Brief, FUE, brand exclusions). /ads meta for the Andromeda + GEM + Lattice era with Entity-ID clustering detection.
• 10-Principle Thinking Framework: every audit, plan, and creative output runs under a shared cognitive discipline. See ads/references/thinking-framework.md.

Full release notes: CHANGELOG.md.

Demo

Sample output

What /ads audit actually returns (truncated for brevity):

{
  "ads_health_score": 73,
  "grade": "C",
  "audit_date": "2026-05-18",
  "platforms": {
    "google_ads": { "score": 78, "grade": "B", "checks_run": 80, "critical": 2, "high": 5 },
    "meta_ads":   { "score": 64, "grade": "C", "checks_run": 50, "critical": 4, "high": 7 },
    "linkedin":   { "score": 81, "grade": "B", "checks_run": 27, "critical": 0, "high": 3 }
  },
  "top_findings": [
    {
      "id": "M-AND-01",
      "severity": "critical",
      "platform": "meta",
      "title": "Andromeda creative similarity > 60%: retrieval suppression risk",
      "impact": "Estimated 20-35% reach loss; 4 ad sets affected",
      "action": "Replace 7 near-duplicate creatives with concept-diverse variants",
      "owner": "creative",
      "eta_days": 7
    },
    {
      "id": "G-AIM-03",
      "severity": "high",
      "platform": "google",
      "title": "AI Max enabled without negative keyword discipline",
      "impact": "Wasted spend ~$1,400/mo on irrelevant queries",
      "action": "Build negative list from 30d search term report",
      "owner": "search",
      "eta_days": 2
    }
  ],
  "quick_wins": [
    "Enable Consent Mode V2 (Privacy Sandbox compatible, ~1 hr)",
    "Pause 3 ad groups failing 3x Kill Rule (saves ~$420/mo)"
  ]
}

Plus a PDF version (/ads report) with health score gauge, platform comparison charts, and zero-overlap layout for client delivery.

Contents

• Installation: 3 ways to add Claude Ads
• Quick Start
• Commands
• Features: what 250+ audit checks cover
• Compared to manual / agency / commercial tools
• Use cases
• Eval harness: verified rigor
• Architecture
• How it analyzes your ads
• FAQ
• Requirements
• Uninstall
• Roadmap
• Project info
• Related projects
• Maintainer
• License

Installation: 3 ways to add Claude Ads

ℹ️ Two install paths — pick the one that matches your access.

The commands below default to the community private mirror at AI-Marketing-Hub/claude-ads (early access for Pro members). To use the public release at AgriciDaniel/claude-ads (MIT, no membership), swap AI-Marketing-Hub/claude-ads for AgriciDaniel/claude-ads and the plugin slug claude-ads@ai-marketing-hub-claude-ads for claude-ads@agricidaniel-claude-ads in every command.

Pro members: requires gh auth login (or PAT) with access to the AI-Marketing-Hub org. If /plugin marketplace add 404s, DM in the Skool community to get added.

1. Plugin install (Claude Code, recommended)

/plugin marketplace add AI-Marketing-Hub/claude-ads
/plugin install claude-ads@ai-marketing-hub-claude-ads

Registers claude-ads as a native plugin with auto-updates, namespace isolation, and version tracking.

2. One-command install (Unix/macOS/Linux)

curl -fsSL https://raw.githubusercontent.com/AI-Marketing-Hub/claude-ads/main/install.sh | bash

2. One-command install (Windows PowerShell)

irm https://raw.githubusercontent.com/AI-Marketing-Hub/claude-ads/main/install.ps1 | iex

3. Cross-host install (Codex CLI / Cursor / Windsurf / Gemini CLI / Goose)

# Unix/macOS/Linux
bash install.sh --target=codex      # OpenAI Codex CLI       (experimental)
bash install.sh --target=cursor     # Cursor IDE              (experimental)
bash install.sh --target=windsurf   # Windsurf IDE            (experimental)
bash install.sh --target=gemini     # Gemini CLI              (experimental)
bash install.sh --target=goose      # Goose CLI               (experimental)

# Windows PowerShell
.\install.ps1 -Target codex
.\install.ps1 -Target cursor
.\install.ps1 -Target windsurf
.\install.ps1 -Target gemini
.\install.ps1 -Target goose

Per-host install path table:

Path overrides:

bash install.sh --target=claude --skill-dir=~/custom/skills --agent-dir=~/custom/agents

Targets and override paths are strictly whitelist-validated: no shell injection, no flag confusion