10 Free Claude Code Skills for Meta Ads

Key Takeaways

• Ten Claude Code skills can fully automate the Meta Ads workflows that agencies typically charge $3,000–$7,000 per month to manage manually.
• Each skill runs as a Claude Code session with a specific system prompt — no external SaaS subscriptions required beyond your Claude and Meta API access.
• The Auto Account Auditor alone surfaces wasted spend patterns that typically represent 15–30% of total ad budget in unoptimized accounts.
• The Bulk Copy Generator uses AIDA and PAS frameworks with Meta character limit enforcement, producing 100+ ad variations in under 3 minutes.
• The Anomaly Detection Bot flags CPM spikes and ROAS drops within 2 standard deviations of the trailing 7-day average — catching problems the same day they appear.
• The Creative Performance Ranker groups ads by angle and format, giving you a data-driven brief for your next production cycle rather than gut-feel decisions.
• Once your winning angles are identified, the AI Creative Prompt Generator feeds them into HeyOz to produce ready-to-upload UGC-style assets — closing the loop from data to creative to live ad.

What Are Claude Code Skills for Meta Ads?

Claude Code skills are reusable system prompts and scripts that turn Claude into a specialized Meta Ads analyst, copywriter, or automation agent. Instead of paying an agency to run manual account audits, write ad copy, or monitor performance anomalies, you define the task once as a Claude Code skill — and run it on demand against your own data.

A skill is essentially a structured prompt that tells Claude what role to play, what data to expect, what analysis to run, and what output format to return. The ten skills in this guide cover every core function a Meta Ads manager handles week-to-week: auditing, pacing, creative analysis, competitor monitoring, anomaly detection, copy generation, custom metric calculation, audience optimization, naming governance, and AI creative briefing.

The global AI in advertising market reached $14.9 billion in 2025 and is projected to exceed $107 billion by 2028. Meanwhile, 72% of performance marketers report that manual campaign management is their biggest time drain . These skills eliminate that drain without a SaaS subscription or an agency retainer.

What Data Do the Skills Use?

Most skills in this guide work with two inputs: (1) a campaign export from Meta Ads Manager in CSV or JSON format, or (2) a direct connection to the Meta Marketing API. Skills 1–3 and 6–10 work entirely from pasted exports — no API integration required. Skills 4 and 5 (Competitor Tracker and Anomaly Bot) optionally connect to the Meta Ad Library API and Marketing API respectively for automated weekly runs.

How Do You Set Up Claude Code for Meta Ads?

Setup takes under 10 minutes. You need Claude Code installed (available on claude.ai/code), a Meta Ads Manager export, and optionally a Meta Marketing API access token for the automated skills.

Step 1: Install Claude Code

Claude Code is Anthropic's official CLI. Install it with npm:

npm install -g @anthropic-ai/claude-code

Or download the desktop app from claude.ai/code. Sign in with your Anthropic account. A Pro plan ($20/month) is sufficient for all ten skills.

Step 2: Create Your Meta Ads Project Folder

mkdir meta-ads-claude

cd meta-ads-claude

touch CLAUDE.md .env

Add your Meta API token to the .env file:

META_ACCESS_TOKEN=your_token_here

META_AD_ACCOUNT_ID=act_your_account_id

META_APP_ID=your_app_id

SLACK_WEBHOOK_URL=your_slack_webhook # optional, for alerts

Step 3: Export Your Campaign Data

In Meta Ads Manager: go to Ads reporting → Customize columns to include Impressions, Reach, Clicks, CTR, CPM, CPC, Spend, ROAS, 3-Second Video Views, ThruPlays, Results, Cost per Result. Export as CSV. Save to your project folder as campaign_export.csv.

Step 4: Create Your CLAUDE.md

This file gives Claude persistent context about your account:

# Meta Ads Account Context

## Account Overview

- Brand: [Your Brand Name]

- Monthly budget: $[X]

- Primary objective: [Conversions / Traffic / Awareness]

- Target CPA: $[X]

- Target ROAS: [X]x

- Primary markets: [US / UK / AU / etc]

## Campaign Structure

- [Campaign 1]: [Description, budget, objective]

- [Campaign 2]: [Description, budget, objective]

## KPIs

- Primary: ROAS (target: [X]x)

- Secondary: CPA (target: $[X])

- Awareness metric: CPM (benchmark: $[X])

## Notes

- [Any account-specific context the agent should know]

What Are the 10 Free Skills and Their Exact Prompts?

Each skill below includes the full system prompt you paste into Claude Code, the expected input, and what the output looks like. g., skill-01-auditor.md)

Skill 1: Auto Account Auditor

The Auto Account Auditor runs 50+ diagnostic checks across your full campaign export and returns a scored health report (0–100) with prioritized action items. It identifies broken pixel events, zero-conversion ad sets, budget cannibalization, ASC audience overlap, and wasted spend patterns — in under 10 minutes.

Input: Paste your full campaign/ad set/ad level CSV export.

Skill Prompt:

You are a senior Meta Ads account auditor with 10+ years of performance marketing experience.

I will paste my campaign export data. Run a comprehensive 50-point account health audit.

Audit structure:

## ACCOUNT HEALTH SCORE: [X/100]

### CRITICAL ISSUES (score: -5 each)

Check for:

- Ad sets with zero conversions in the last 30 days and spend > $50

- Campaigns with frequency > 4.0 (audience fatigue)

- Pixel firing errors (if conversion data shows gaps vs spend)

- Ads in 'LIMITED' delivery status with no reason given

- Budget allocation: single ad set getting >80% of campaign budget (starvation)

### HIGH PRIORITY (score: -3 each)

- CTR below 0.8% on traffic/conversion campaigns

- CPM above 3x account average

- Ad sets with audience overlap >30% (use Meta's overlap tool logic)

- No creative testing: campaigns running same creative for 30+ days

- Advantage+ Shopping campaigns with <5 active creatives

- Retargeting windows: check if 180-day window is pulling in cold audiences

### MEDIUM PRIORITY (score: -1 each)

- Campaign names don't follow convention (can't segment by funnel stage)

- No ad schedule set on awareness campaigns (running 24/7 vs peak hours)

- Single interest targeting without broad expansion test

- Video ads with <15% hook rate (3-sec views / impressions)

- Missing UTM parameters on destination URLs

### WASTED SPEND ANALYSIS

Calculate:

- Spend on zero-conversion ad sets: $[X] ([Y]% of total)

- Spend on high-frequency creatives (freq >4): $[X]

- Estimated recoverable budget per month: $[X]

### TOP 5 IMMEDIATE ACTIONS

Prioritize by impact/effort ratio. Be specific (campaign name, ad set name, exact action).

### BUDGET REALLOCATION RECOMMENDATION

Based on ROAS by campaign, recommend specific budget shifts.

Data: [PASTE CSV HERE]

Skill 2: Budget Pacing Sentinel

The Budget Pacing Sentinel takes your daily spend data and projects end-of-month delivery. It flags any account or campaign pacing more than 15% over or under target and generates a Slack-ready alert message. Overpacing is one of the top three budget management mistakes in paid social — this skill prevents it.

Input: Daily spend breakdown (account/campaign level) plus monthly budget targets.

Skill Prompt:

You are a Meta Ads budget pacing analyst. I will give you daily spend data and monthly budgets.

Analyze pacing and return a complete pacing report.

For each campaign:

1. Calculate: days elapsed / days in month = expected spend %

2. Calculate: actual spend / monthly budget = actual spend %

3. Pacing variance = actual % - expected %

4. Flag if variance > +15% (overpacing) or < -15% (underpacing)

5. Project end-of-month spend at current daily run rate

6. Flag if projected spend > 110% or < 90% of monthly budget

Output format:

## BUDGET PACING REPORT — [Date]

### ACCOUNT OVERVIEW

- Today is Day [X] of [Y] (expected spend: [Z]%)

- Total account spend to date: $[X] / $[Y] monthly budget ([Z]%)

- Projected EOM spend: $[X] ([Y]% of budget)

- Pacing status: [ON TRACK / OVERPACING / UNDERPACING]

### CAMPAIGN PACING TABLE

| Campaign | Budget | Spent | Expected% | Actual% | Variance | Projected EOM | Status |

### ALERTS

For each flagged campaign:

🔴 OVERPACING: [Campaign Name] — at current rate, will spend $[X] ($[Y] over budget)

Recommended action: Reduce daily budget to $[X] to normalize

🟡 UNDERPACING: [Campaign Name] — projected to underspend by $[X]

Recommended action: Increase daily budget to $[X] or extend end date

### SLACK ALERT MESSAGE (copy-paste ready)

:meta: *Budget Pacing Alert — [Date]*

Accounts flagged: [N]

[For each alert: emoji + campaign + status + action]

Data:

- Today's date: [DATE]

- Monthly budgets: [PASTE BUDGET TABLE]

- Daily spend data: [PASTE SPEND DATA]

Skill 3: Creative Performance Ranker

The Creative Performance Ranker pulls ad-level data and ranks every creative by ROAS and CTR. It then groups results by creative angle (problem/solution, social proof, product demo, testimonial, UGC) and format (static image, short video, carousel) so you know exactly what to produce more of.

Skill Prompt:

You are a creative strategist analyzing Meta Ads performance data.

I will paste ad-level performance data. Rank all ads and identify winning angles.

Step 1 — Tag each ad with:

- Format: [Static Image | Short Video (<15s) | Long Video (15s+) | Carousel | Collection]

- Angle: [Problem/Solution | Social Proof | Product Demo | Testimonial | UGC Style |

Offer/Discount | Before/After | Education | Humor | Urgency]

(Infer angle from ad name if possible; ask if ambiguous)

Step 2 — Rank all ads by ROAS (descending), then by CTR as tiebreaker

Minimum spend threshold for ranking: $50 (exclude ads with less spend)

Step 3 — Generate angle-level analysis:

For each angle that has 2+ ads:

- Average ROAS

- Average CTR

- Average Hook Rate (3-sec views / impressions — if data available)

- Total spend

- Winner count (ads in top 25% of ROAS)

Step 4 — Format-level analysis (same metrics)

Step 5 — Creative brief recommendation:

Top 3 winning angle+format combinations with brief specs:

"Your data shows [ANGLE] + [FORMAT] generates [X]x ROAS vs [Y]x account average.

Brief: [specific creative direction based on winning patterns]"

### CREATIVE SCORECARD

| Rank | Ad Name | Format | Angle | ROAS | CTR | Hook Rate | Spend | Score |

### WINNING COMBINATIONS

### LOSING COMBINATIONS (kill/pause these)

### NEXT SPRINT BRIEF

Ad data: [PASTE AD-LEVEL EXPORT]

Skill 4: Competitor Ad Tracker

The Competitor Ad Tracker queries the Meta Ad Library API to surface new creatives your competitors launched in the past 7 days, spots patterns in their messaging, and alerts you when they scale spend on a particular angle. You stop briefing reactively and start briefing proactively.

Setup Script (run weekly via cron):

#!/usr/bin/env python3

# competitor_tracker.py — runs weekly, feeds output to Claude for analysis

import requests, json, os

from datetime import datetime, timedelta

from dotenv import load_dotenv

load_dotenv()

ACCESS_TOKEN = os.getenv('META_ACCESS_TOKEN')

COMPETITORS = [

{"name": "Competitor A", "page_id": "PAGE_ID_HERE"},

{"name": "Competitor B", "page_id": "PAGE_ID_HERE"},

]

def fetch_ads(page_id, days_back=7):

since = (datetime.now() - timedelta(days=days_back)).strftime('%Y-%m-%d')

url = f"https://graph.facebook.com/v21.0/ads_archive"

params = {

"access_token": ACCESS_TOKEN,

"ad_reached_countries": '["US"]',

"search_page_ids": f'["{page_id}"]',

"ad_delivery_date_min": since,

"fields": "id,ad_creative_bodies,ad_creative_link_captions,ad_creative_link_descriptions,ad_creative_link_titles,ad_snapshot_url,ad_delivery_start_time,impressions",

"limit": 50

}

r = requests.get(url, params=params)

return r.json().get('data', [])

results = {}

for comp in COMPETITORS:

results[comp['name']] = fetch_ads(comp['page_id'])

with open('competitor_ads.json', 'w') as f:

json.dump(results, f, indent=2)

print(f"Fetched ads for {len(COMPETITORS)} competitors. Saved to competitor_ads.json")

Skill Prompt (after running script):

You are a competitive intelligence analyst for a Meta Ads team.

I will paste JSON data from the Meta Ad Library showing my competitors' active ads from the past 7 days.

Analyze and return a competitive intelligence brief.

## COMPETITIVE INTEL BRIEF — Week of [DATE]

### NEW ADS LAUNCHED (past 7 days)

For each competitor:

- [Competitor Name]: [N] new ads

- Top format: [format with count]

- Primary angle: [dominant message theme]

- New creative hook: "[notable opening line or visual concept]"

- Estimated impression range: [from API data]

### MESSAGING PATTERN ANALYSIS

For each competitor, identify:

- Core value proposition (what do they lead with?)

- Proof elements (testimonials, stats, social proof)

- CTA language (what action are they pushing?)

- Offer strategy (discount, free trial, free shipping, etc.)

### EMERGING TRENDS

What themes appear across multiple competitors this week?

What angles are being scaled (high impression volume)?

What angles appear to be tests (low impression, new launch)?

### GAPS IN COMPETITOR MESSAGING

What angles or proof elements are your competitors NOT using that represent an opportunity?

### RECOMMENDED RESPONSE BRIEF

Top 2-3 creative briefs to counter competitor activity, with specific hooks and angles.

Data: [PASTE competitor_ads.json content]

Skill 5: Anomaly Detection Bot

The Anomaly Detection Bot compares today's performance against your trailing 7-day average and flags any metric that has moved more than 2 standard deviations. It generates a same-day Slack alert with the exact campaign, metric, and magnitude of the shift — so you catch creative fatigue, pixel fires, or budget issues before they compound.

Skill Prompt:

You are a Meta Ads anomaly detection system. I will paste daily performance data for the last 8 days

(7 days baseline + today). Identify statistical anomalies and generate alerts.

For each campaign and key metric, calculate:

- 7-day mean

- 7-day standard deviation

- Today's value

- Z-score: (today - mean) / std_dev

- Flag if |z-score| > 2.0

Metrics to check:

- CPM (spike = bad)

- CTR (drop = bad)

- CPC (spike = bad)

- ROAS (drop = bad)

- Frequency (spike = fatigue signal)

- Conversion rate (drop = landing page or pixel issue)

- Spend (spike or drop = budget or delivery issue)

## ANOMALY DETECTION REPORT — [Date]

### CRITICAL ANOMALIES (z-score > 3.0)

🚨 [Campaign Name] — [Metric]: Today [X] vs 7-day avg [Y] (z-score: [Z])

Possible cause: [most likely explanation]

Recommended action: [specific next step]

### WARNING ANOMALIES (z-score 2.0–3.0)

⚠️ [Campaign Name] — [Metric]: Today [X] vs 7-day avg [Y] (z-score: [Z])

### NORMAL CAMPAIGNS

✅ [list campaigns with no anomalies]

### SLACK ALERT (copy-paste)

:rotating_light: *Meta Ads Anomaly Alert — [Date]*

Critical: [N] | Warning: [N]

[For each critical: campaign + metric + magnitude + action]

Performance data (last 8 days by campaign):

[PASTE DATA]

Skill 6: Bulk Copy Generator

The Bulk Copy Generator produces 100+ headline and body copy variations using AIDA (Attention, Interest, Desire, Action) and PAS (Problem, Agitate, Solution) frameworks. It enforces Meta's character limits automatically — 40 characters for headlines, 125 for primary text — and organizes output by angle so your designer can match copy to creative.

Skill Prompt:

You are a direct response copywriter specializing in Meta Ads. Generate 100+ ad copy variations.

Product/Brand context:

- Product: [DESCRIBE YOUR PRODUCT]

- Target audience: [WHO YOU'RE TARGETING]

- Primary benefit: [MAIN VALUE PROP]

- Secondary benefits: [LIST 3-5]

- Social proof: [TESTIMONIALS/STATS/REVIEWS YOU HAVE]

- Offer: [CURRENT PROMOTION OR CTA]

- Top objections: [WHAT PREVENTS PEOPLE FROM BUYING]

- Brand voice: [TONE — e.g., direct, conversational, premium]

Generate copy organized by framework and angle.

Meta character limits (ENFORCE STRICTLY):

- Headline: max 40 characters

- Primary text: max 125 characters (first line — most important)

- Description: max 30 characters

### AIDA FRAMEWORK (50 variations)

#### Problem-Led (10 headlines + 10 primary texts)

#### Benefit-Led (10 headlines + 10 primary texts)

#### Curiosity-Led (10 headlines + 10 primary texts)

#### Social Proof-Led (10 headlines + 10 primary texts)

#### Offer-Led (10 headlines + 10 primary texts)

### PAS FRAMEWORK (50 variations)

#### Fear/Risk (10 variations: headline + body paired)

#### Frustration (10 variations)

#### Aspiration (10 variations)

#### Transformation (10 variations)

#### Authority (10 variations)

For each variation format as:

H: [headline — max 40 chars] ([char count])

P: [primary text — max 125 chars] ([char count])

Flag any that exceed limits with ⚠️

Skill 7: Custom Metrics Calculator

Meta Ads Manager does not surface Hook Rate or Hold Rate by default, despite these being the two most predictive metrics for video ad performance. Hook Rate (3-second views ÷ impressions) tells you if your opening grabs attention. Hold Rate (ThruPlays ÷ 3-second views) tells you if people stay after the hook. This skill calculates both — plus a dozen other derived metrics — from your standard export.

Skill Prompt:

You are a Meta Ads data analyst. I will paste ad-level performance data.

Calculate custom metrics that Meta Ads Manager doesn't show natively.

For each ad, calculate these custom metrics:

## VIDEO METRICS (only for video ads)

- Hook Rate = 3-Second Video Views / Impressions × 100

Benchmark: >30% = strong, 15-30% = average, <15% = weak hook

- Hold Rate = ThruPlays / 3-Second Video Views × 100

Benchmark: >25% = strong, 10-25% = average, <10% = loses people fast

- Video Efficiency Score = Hook Rate × Hold Rate / 100

(composite score — higher = video is both grabbing and retaining)

## ENGAGEMENT METRICS (all ad types)

- Engagement Rate = (Clicks + Post Reactions + Comments + Shares) / Impressions × 100

- Scroll Stop Rate = Link Clicks / Impressions × 100 (proxy for scroll-stopping power)

- Click Quality Score = Conversions / Clicks × 100 (landing page effectiveness proxy)

## EFFICIENCY METRICS

- Effective CPM (eCPM) = Spend / Impressions × 1000

- Cost per Engaged User = Spend / (Post Reactions + Comments + Shares)

- True CPA = Spend / All Conversion Events (not just primary)

## CREATIVE FATIGUE INDICATORS

- Frequency-adjusted CTR: CTR normalized for frequency (CTR / (1 + (Frequency-1) × 0.1))

Drop in this metric = fatigue signal even if raw CTR holds

Output as a table with color-coding indicators:

✅ Above benchmark | ⚠️ At benchmark | 🔴 Below benchmark

Then: Flag top 3 ads by Video Efficiency Score and bottom 3 as priority kills.

Data: [PASTE AD-LEVEL EXPORT WITH VIDEO METRICS]

Skill 8: Retargeting Window Optimizer

The Retargeting Window Optimizer analyzes conversion performance across your 7-day, 14-day, and 30-day retargeting windows to identify where you're overspending on cold audiences masquerading as warm ones. It recommends window adjustments based on your actual customer purchase cycle — not Meta's defaults.

Skill Prompt:

You are a Meta Ads audience strategist. I will paste performance data segmented by retargeting window.

Analyze which windows convert and recommend budget allocation.

Input data should include performance for these audience segments (or as many as you have):

- Website visitors: 7-day, 14-day, 30-day, 60-day, 90-day, 180-day

- Video viewers: 25%, 50%, 75%, 95% watch time segments

- Engaged with page: 7-day, 30-day, 90-day

- Add to cart: 7-day, 14-day, 30-day

- Initiate checkout: 7-day, 14-day

For each segment, analyze:

- Conversion rate relative to account average

- CPA relative to account average

- ROAS relative to account average

- Audience size (estimate from impressions at similar CPM)

- Spend efficiency score: ROAS / (CPM / account_avg_CPM)

## RETARGETING WINDOW ANALYSIS

### HIGHEST VALUE WINDOWS (ROAS > account average)

[List with metrics]

### LOWEST VALUE WINDOWS (ROAS < account average)

[List — these are likely polluted with cold audience]

### AUDIENCE OVERLAP RISK

Flag windows where size suggests overlap (e.g., 180-day window contains 7-day window).

Recommend exclusion stacking: always exclude 7-day from 30-day, etc.

### RECOMMENDED WINDOW STRUCTURE

Based on data:

- Tier 1 (hottest): [window] — suggested budget: [X]% of retargeting budget

- Tier 2 (warm): [window] — suggested budget: [X]%

- Tier 3 (cool): [window] — suggested budget: [X]%

- Pause: [windows to pause]

### BUDGET REALLOCATION

Current estimated waste: $[X]/month on underperforming windows

Recommended shift: Move $[X] from [window] to [window]

Data: [PASTE AUDIENCE-SEGMENTED PERFORMANCE DATA]

Skill 9: Campaign Naming Validator

Inconsistent campaign naming destroys your ability to segment creative performance in reporting. The Campaign Naming Validator enforces a standard convention across your account, catches violations, and outputs a corrected naming sheet you can bulk-upload. Without naming standards, you cannot reliably group performance by funnel stage, audience type, or creative format.

Recommended Naming Convention:

CAMPAIGN: [BRAND]_[OBJECTIVE]_[FUNNEL]_[AUDIENCE]_[DATE]

Example: BRANDX_CONV_TOFU_BROAD_2026-05

AD SET: [AUDIENCE_TYPE]_[WINDOW/SIZE]_[PLACEMENT]_[DATE]

Example: RETARG_7D_FEED_2026-05

AD: [FORMAT]_[ANGLE]_[VERSION]_[DATE]

Example: VID_SOCIAL-PROOF_V3_2026-05

Abbreviations:

Objective: CONV, TRAF, AWA, LEAD, CATA

Funnel: TOFU, MOFU, BOFU

Format: VID (video), IMG (static image), CAR (carousel), COL (collection)

Angle: PROB (problem), SOC (social proof), DEM (demo), OFF (offer), UGC, BEF (before/after)

Skill Prompt:

You are a Meta Ads account structure specialist. I will paste a list of campaign, ad set, and ad names.

Validate against our naming convention and output a corrected naming sheet.

Naming convention:

Campaign: [BRAND]_[OBJECTIVE]_[FUNNEL]_[AUDIENCE]_[YYYY-MM]

Ad Set: [AUDIENCE_TYPE]_[WINDOW]_[PLACEMENT]_[YYYY-MM]

Ad: [FORMAT]_[ANGLE]_[VERSION]_[YYYY-MM]

For each entity:

1. Parse the current name

2. Identify which convention elements are present / missing / wrong

3. Generate a corrected name following the convention

4. Flag ambiguous cases where you cannot infer the correct value

## NAMING VALIDATION REPORT

### COMPLIANCE SCORE: [X]% of [N] entities are convention-compliant

### VIOLATIONS TABLE

| Current Name | Issue | Corrected Name | Confidence |

### PATTERNS IN VIOLATIONS

[What naming habits are most common — e.g., 'missing funnel stage' or 'date format inconsistent']

### BULK RENAME SHEET (tab-separated, import to rename tool)

OLD NAME | NEW NAME

### AMBIGUOUS CASES — NEED YOUR INPUT

[Cases where you cannot infer the correct value — ask for clarification]

Entity list: [PASTE CAMPAIGN / AD SET / AD NAMES]

Skill 10: AI Creative Prompt Generator

The AI Creative Prompt Generator reads your winning angles from the Creative Performance Ranker output and produces production-ready prompts for both image and video ads. These prompts are formatted specifically for AI generation tools — so you go directly from performance data to a fully briefed creative asset, with no intermediate brief writing.

Once the prompts are generated, run them through HeyOz to get actual AI-generated UGC and video ad assets back. This closes the full loop: performance data → winning angles → AI prompts → AI-generated creatives → Meta upload.

Skill Prompt:

You are an AI creative director. I will paste the output from my Creative Performance Ranker.

Generate production-ready prompts for AI image and video ad generation.

For each of the top 3 winning angle+format combinations, generate:

### IMAGE AD PROMPTS (for tools like HeyOz, Midjourney, DALL-E)

Prompt structure:

"[SUBJECT/SCENE]. [PRODUCT CONTEXT]. [LIGHTING AND MOOD]. [STYLE]. [COMPOSITION]. [TEXT OVERLAY IF ANY]. [NEGATIVE PROMPTS]."

For each winning angle, generate:

1. Primary image prompt (hero creative, 1:1 ratio)

2. Story variant (9:16 ratio, vertical)

3. Feed variant (4:5 ratio)

Include:

- Specific visual metaphors that match the winning angle

- Realistic, UGC-style direction (avoid stock photo aesthetics)

- Text overlay placement guidance (CTA position, headline position)

- Brand color references if relevant

### VIDEO AD SCRIPTS (for UGC-style video generators)

For each winning angle, generate a 15-second and 30-second script:

Format:

[0-3s HOOK]: [Exact words or visual action — must match Hook Rate data]

[3-8s PROBLEM/TENSION]: [Build the pain point or desire]

[8-12s SOLUTION]: [Introduce product/benefit]

[12-15s CTA]: [Exact CTA with offer if applicable]

[30s version extends each section proportionally]

Also generate:

- On-screen text for each segment (captions/supers)

- B-roll description for each section

- UGC persona brief: age, style, location, tone

### HEYOZ GENERATION BRIEF

For each set of prompts, include a one-paragraph brief for running through HeyOz:

"Use [format type] with [specific settings]. The winning angle is [X] — ensure the hook

establishes [Y] within the first 2 seconds. Reference imagery: [visual direction]."

Performance ranker output: [PASTE CREATIVE PERFORMANCE RANKER OUTPUT]

How Do You Build a Full Meta Ads Automation Stack?

The ten skills above work as a coordinated stack — not as isolated tools. Here is how they connect in a weekly operational cadence:

Weekly Workflow (2 hours vs. 20 hours manual)

• Monday (15 min): Run Skill 1 (Account Auditor) on fresh weekly export. Implement top 3 actions from the report.
• Monday (5 min): Run Skill 9 (Naming Validator) on any new campaigns launched last week.
• Daily (5 min): Skill 2 (Pacing Sentinel) and Skill 5 (Anomaly Bot) run automatically via cron and push to Slack — only requires action if alerts fire.
• Wednesday (20 min): Run Skill 3 (Creative Ranker) and Skill 7 (Custom Metrics) together to get the full creative performance picture. Feed output into Skill 10 (Prompt Generator).
• Wednesday (10 min): Take Skill 10 output into HeyOz to generate 3-5 new creatives based on the winning angles.
• Thursday (15 min): Run Skill 4 (Competitor Tracker) to check for new competitor ads. Brief any counter-creative needed.
• Friday (20 min): Run Skill 6 (Copy Generator) to produce fresh copy for next week's test set. Run Skill 8 (Retargeting Optimizer) if launching new retargeting audiences.

What Does This Replace?

A mid-market Meta Ads agency typically charges $3,000–$7,000 per month for: weekly reporting, creative briefs, copy writing, competitive monitoring, and performance analysis. The Heyoz SEO Agency approach applies the same logic to content marketing — systematizing what was previously bespoke agency work into repeatable, AI-powered workflows.

These ten skills cover the analytical and copy functions. The creative production function — generating actual ad assets — is handled by HeyOz. Paste a product URL into HeyOz, select the angle and format from your Skill 10 output, and get production-ready ad creatives in minutes.

Comparison: Agency vs. Claude Code Stack

Agency vs. Claude Code Stack

Account audit: Agency = Monthly, 4–8 hrs. Claude Code = Skill 1 — weekly, 10 min. Time saved = ~2 hrs/week.

Budget monitoring: Agency = Manual daily check. Claude Code = Skill 2 — automated alert. Time saved = ~3 hrs/week.

Creative analysis: Agency = Weekly, 3–4 hrs. Claude Code = Skills 3 + 7, 20 min. Time saved = ~3 hrs/week.

Competitor monitoring: Agency = Ad hoc, 2–3 hrs. Claude Code = Skill 4 — automated weekly. Time saved = ~2 hrs/week.

Copy writing: Agency = Per sprint, 4–6 hrs. Claude Code = Skill 6 — 3 min. Time saved = ~4 hrs/week.

Creative briefing: Agency = Per sprint, 2–3 hrs. Claude Code = Skill 10 + HeyOz, 30 min. Time saved = ~2 hrs/week.

Total time saved: approximately 16+ hours per week. At a $75/hr opportunity cost, that is $1,200/week or $4,800/month — close to what agencies charge for the same scope.

Frequently Asked Questions

Do these Claude Code skills require a paid Claude plan?

Claude Pro ($20/month) is sufficient for all ten skills. Skills that process large CSV exports (Audit, Creative Ranker) will use more context window — stick to exports of 200 ads or fewer per session on Pro, or upgrade to Claude Max for larger accounts.

Do Skills 4 and 5 require Meta API access?

Skill 4 (Competitor Tracker) uses the Meta Ad Library API, which is free and public — you only need a Meta developer account and a basic access token. Skill 5 (Anomaly Bot) can work with manual daily exports (no API needed) or connect to the Marketing API for automated daily runs. Both API setups take under 30 minutes.

How accurate is the Account Auditor compared to a human audit?

The Auto Account Auditor covers structural and data-driven checks with high accuracy — it will reliably catch zero-conversion ad sets, frequency issues, budget starvation, and CTR anomalies. Human auditors add value in areas Claude cannot see without additional context: account history, seasonal patterns, and brand strategy decisions. Use the skill as a first-pass filter, then apply judgment to the top 5 recommendations.

Can I automate these skills to run on a schedule?

Yes. Skills 2 and 5 (Pacing Sentinel and Anomaly Bot) are designed for daily automation via cron jobs or GitHub Actions. The Python scripts included in Skills 2, 4, and 5 can be scheduled to pull data from the API, feed it to Claude via the Anthropic API, and push alerts to Slack — fully headless, no manual input required.

What is the full cost of this stack vs. an agency?

Claude Pro: $20/month. Meta API access: free. Python hosting for automated scripts: ~$5–10/month on a basic VPS or free on GitHub Actions. HeyOz for creative production: $44.99/month. Total stack cost: under $80/month. Agency equivalent: $3,000–$7,000/month for the same scope of work.

How do I integrate Skill 10 output with HeyOz?

Skill 10 generates structured prompts with a HeyOz Generation Brief section. Copy the brief, go to heyoz.com, paste your product URL, and use the generated prompt as your creative direction input. HeyOz handles the actual image or video generation, editing, and format resizing — you receive production-ready assets for Meta upload.