10 Free Claude Skills That Replace Your Meta Ads Reporting Stack

Key Takeaways

• 10 free, open-source Claude skills turn CSV exports from Meta Ads Manager into actionable analysis — creative fatigue scanning, wasted spend detection, CPA diagnosis, and more.

• No API keys, no MCP setup, no coding required. Export a CSV, drop it into Claude, and start asking questions.

• Two setup paths: Claude Projects (browser only, 10 minutes) or Claude Code (more powerful, 15-20 minutes). Both work with the same CSV exports.

• All skills come from four free GitHub repositories. Three repos give you full coverage of all 10 skills.

• Your data stays local — no third-party ad analytics service ever sees your data. Total cost is $20/month for a Claude Pro subscription.

• Skills cover the full media buyer workflow: fatigue detection, waste auditing, A/B test significance, budget modeling, weekly reporting, ad copy generation, competitor analysis, and lookalike audience auditing.

Introduction

No API keys. No MCP setup. No coding. Just export a CSV from Ads Manager, drop it in, and start asking questions.

Most marketers pay $200-500/month for ad analytics tools that do what a free Claude skill can do in 30 seconds. The difference: these skills run entirely on your machine. Your data never touches a third-party server beyond Claude itself.

This guide walks you through 10 free skills for analyzing Meta Ads data with Claude, where to get them, and how to set everything up from scratch — even if you have never opened a terminal.

What Are Claude Skills (and Why Should You Care)?

A Claude "skill" is just a text file — a set of instructions written in plain English — that teaches Claude how to do a specific job. When you give Claude a skill file for Meta Ads analysis, it knows exactly what metrics to look at, what benchmarks to compare against, and what recommendations to make.

Think of it like hiring a specialist. Without a skill, Claude is a smart generalist. With a Meta Ads skill loaded, it becomes a media buying analyst who knows platform benchmarks, auction mechanics, creative fatigue patterns, and budget optimization frameworks.

There are two ways to use these skills:

• Claude Projects (easiest, no terminal needed): Upload the skill files as "knowledge" in a Claude Project at claude.ai. Works on any computer with a browser.

• Claude Code (more powerful): Install the skill files locally so Claude Code can use them as slash commands. Requires the Claude Code app but unlocks more advanced features.

Both methods work with the same CSV exports from Meta Ads Manager. We will cover both setup paths below.

The 10 Skills (and What Each One Does)

1. Creative Fatigue Scanner

What it does: Analyzes every active ad for signs of creative exhaustion. Tracks CTR decay over time, frequency buildup, and CPC inflation. Categorizes each ad as healthy (keep running), warning (refresh soon), or critical (pause now).

Why it matters: Creative fatigue is the number one silent ROAS killer. Most advertisers do not catch it until performance has already cratered. This skill flags the early warning signs — a 15% CTR drop over 7 days, frequency crossing 3.0, or CPC climbing while impressions hold steady.

What to export from Ads Manager: Ad-level data with these columns: Ad Name, Impressions, Reach, Frequency, CTR (Link), CPC, Amount Spent, Results, Date (daily breakdown for at least 14 days).

Example prompt once loaded:

"Analyze this CSV for creative fatigue. Flag any ad where CTR has declined more than 15% over the last 14 days or frequency is above 3. Categorize each as healthy, warning, or critical."

2. Wasted Spend Finder

What it does: Scans your entire account for ads and ad sets producing zero conversions — or conversions at a CPA so high they are effectively wasted. Groups the waste by campaign, ad set, or creative theme and attaches dollar amounts.

Why it matters: Most Meta ad accounts have 15-25% hidden waste — budget flowing to placements, audiences, or creatives that have never converted and likely never will. This skill quantifies exactly where that money is going.

What to export: Campaign, Ad Set, and Ad-level data with: Name fields, Amount Spent, Results (Purchases or your conversion event), Cost per Result, Impressions, Link Clicks.

Example prompt:

"Find every ad set and ad that spent more than $50 in the last 30 days with zero purchases. Group them by campaign and show total wasted spend per group."

3. CPA Spike Diagnosis

What it does: When your CPA suddenly jumps, this skill isolates the root cause. It compares current vs. previous period data and diagnoses whether the spike came from audience fatigue, creative decay, bid competition, budget scaling too fast, or external factors (seasonality, algorithm changes).

Why it matters: A CPA spike can have five different root causes, each requiring a different fix. Pausing ads when the real problem is audience saturation wastes the creative. Refreshing creative when the real problem is bid competition wastes the effort.

What to export: Two date ranges — your "spike" period and a baseline period. Same columns: Campaign Name, Ad Set Name, Amount Spent, Impressions, Reach, Frequency, CTR, CPC, CPM, Results, Cost per Result.

Example prompt:

"Compare these two CSV exports — March 1-15 (baseline) vs March 16-31 (spike period). My CPA jumped 40%. Diagnose the most likely root causes, ranked by dollar impact."

4. Audience Overlap Detector

What it does: Maps every ad set's targeting against every other ad set in your account. Flags where you are bidding against yourself in the same auction — driving up your own CPMs.

Why it matters: Audience overlap is invisible in Ads Manager's default views. Two ad sets can look like they have completely different targeting but share 60% of the same users. You are paying Meta twice to reach the same person — and competing against yourself in the auction.

What to export: Ad Set-level data with: Ad Set Name, Targeting (description), Reach, Impressions, CPM, Amount Spent, Frequency. Also helpful: go to Ads Manager > Ad Sets > select multiple > "Inspect" > "Audience Overlap" for the raw overlap percentages.

Example prompt:

"Map the audience overlap between all my active ad sets. Flag any pairs with more than 30% overlap. Show the CPM difference and estimated wasted spend from self-competition."

5. A/B Test Analyzer

What it does: Takes your split test data and calculates statistical significance. Tells you whether your "winner" is actually a winner — or whether you are calling the test too early with insufficient data.

Why it matters: Most marketers call A/B tests after 2-3 days based on gut feel. A proper significance test requires enough conversions to rule out random chance. This skill does the math and tells you: confident winner, likely winner (needs more data), or no meaningful difference.

What to export: Ad-level data for your test variants: Ad Name, Impressions, Link Clicks, CTR, Results, Cost per Result, Amount Spent. Include at least 7 days of data.

Example prompt:

"Run a statistical significance test on these two ad variants. Tell me if we have a winner at 95% confidence, or how many more conversions we need before we can call it."

6. Budget Reallocation Modeler

What it does: Models what would happen to your CPA at different spend levels using your actual historical performance data. Shows where scaling works (CPA stays flat), where diminishing returns begin, and where returns collapse entirely.

Why it matters: "Just increase the budget" is not a strategy. Every audience and creative has a saturation point where additional spend produces rapidly diminishing returns. This skill maps that curve using your real data instead of guessing.

What to export: Campaign or Ad Set-level data with daily breakdown over at least 30 days: Name, Date, Amount Spent, Results, Cost per Result, Impressions, Reach, Frequency.

Example prompt:

"Model the relationship between daily spend and CPA for each of my top 5 campaigns over the last 30 days. Show me where diminishing returns start and recommend optimal daily budgets."

7. Weekly Performance Report

What it does: Takes this week's raw numbers and generates a complete executive summary: campaign performance table, creative rankings, audience insights, spend pacing, and a prioritized action list. Plain English, no jargon, ready to send to a client or manager.

Why it matters: Writing weekly reports takes 1-2 hours. This skill does it in 30 seconds. The output is formatted, contextualized, and includes specific recommendations — not just a data dump.

What to export: All active campaigns, ad set and ad-level data for the current week: all standard performance columns plus any conversion events you track.

Example prompt:

"Generate a weekly performance report from this CSV. Include an executive summary, campaign ranking table, top and bottom 3 creatives, and a prioritized list of 5 actions for next week."

8. Ad Copy Generator

What it does: Takes your product brief and audience data, then generates 20+ copy variations across Meta's placement specs — Feed, Stories, Reels — with correct character limits for each. Outputs primary text, headlines, and descriptions in short, medium, and long-form variants.

Why it matters: Creative testing at scale requires volume. Writing 20 variations manually takes hours. This skill generates them in seconds, calibrated to your specific product positioning and audience language.

What to provide: Your product URL or brief, target audience description, and optionally your top-performing ad copy for tone matching. No CSV needed for this one — it is generative, not analytical.

Example prompt:

"Generate 20 ad copy variations for [product]. Target audience: [description]. Include 5 short-form (Feed), 5 medium-form (Stories), 5 long-form (Feed), and 5 hook variations for Reels. Follow Meta's character limits."

9. Competitor Ad Library Analyzer

What it does: You give it a competitor name, and it pulls insights from Meta's Ad Library — categorizing their active ads by theme, identifying their top hooks and angles, and suggesting what you should test against them.

Why it matters: Your competitors are already testing messaging that works in your market. Analyzing their ad library systematically (instead of scrolling through it manually) reveals patterns: what hooks they repeat, what offers they push, what formats they favor.

How to use: Go to facebook.com/ads/library, search for your competitor, screenshot or copy their active ads. Paste the data into Claude alongside this skill.

Example prompt:

"Here are 15 active ads from [competitor] pulled from the Meta Ad Library. Categorize them by creative theme and messaging angle. Identify their top 3 hooks and suggest 5 counter-angles I should test."

10. Lookalike Audience Refresher

What it does: Audits your current lookalike audience performance over time. Flags which ones have degraded (lookalikes decay every 60-90 days as the source data ages), recommends which seed lists to update, what percentage to test, and which to retire.

Why it matters: Lookalike audiences are not "set and forget." The underlying seed audience changes as your customer base evolves. A 1% lookalike from 6 months ago may now be targeting a very different profile than your current best customers.

What to export: Ad Set-level data with daily breakdown over 90 days for all ad sets using lookalike audiences: Ad Set Name, Audience Type, Date, Impressions, Reach, CPM, Results, Cost per Result, Frequency.

Example prompt:

"Audit these lookalike audiences over the last 90 days. Flag any where CPA has increased more than 25% from the first 30 days to the last 30 days. Recommend which seed lists to refresh and what percentages to test."

How to Set Up Each Skill (Skill-by-Skill Guide)

Here is exactly which repo each skill comes from, which file to grab, and how to get it running. We have listed both the Claude Projects method (browser only) and the Claude Code method (terminal) for each one.

1. Creative Fatigue Scanner

Source repo: irinabuht12-oss/marketing-skills

File: Skills for Claude/Creative Fatigue Detection.md

Claude Projects setup:

• Go to github.com/irinabuht12-oss/marketing-skills

• Click "Code" > "Download ZIP" and extract it

• Find Skills for Claude/Creative Fatigue Detection.md in the extracted folder

• Go to claude.ai > Projects > your Meta Ads project > "Add knowledge" > upload the file

• Upload your CSV and ask Claude to scan for fatigue

Claude Code setup:

• Download the same file

• Copy it to ~/.claude/skills/creative-fatigue/SKILL.md

• Open Claude Code, drop your CSV, and ask for a creative fatigue scan

Also available in: AgriciDaniel/claude-ads (run /ads audit — creative fatigue is part of the full audit)

2. Wasted Spend Finder

Source repo: AgriciDaniel/claude-ads

Included in: The core /ads audit command (wasted spend detection is one of the 186 checks)

Claude Code setup (recommended):

• Open your terminal and run: curl -fsSL https://raw.githubusercontent.com/AgriciDaniel/claude-ads/main/install.sh | bash

• Open Claude Code, drop your CSV, and run /ads audit

• The audit flags zero-conversion spend automatically as part of the full report

Claude Projects setup:

• Go to github.com/AgriciDaniel/claude-ads

• Download the repo ZIP and extract it

• Find ads/references/meta-audit.md — this contains the audit framework including wasted spend checks

• Upload it to your Claude Project as knowledge

• Upload your CSV and ask: "Find all ad sets and ads with zero conversions and calculate total wasted spend"

3. CPA Spike Diagnosis

Source repo: mathiaschu/meta-ads-analyzer

Files: skill/references/performance_fluctuations.md + skill/references/breakdown_effect.md

Claude Code setup (recommended):

• Go to github.com/mathiaschu/meta-ads-analyzer, download the ZIP

• Copy the entire skill/ folder to ~/.claude/skills/meta-ads-analyzer/

• On Mac: Finder > Cmd+Shift+G > paste ~/.claude/skills/ > create meta-ads-analyzer folder > drop in the files

• On Windows: navigate to %USERPROFILE%\.claude\skills\ > same process

• Open Claude Code with two CSV exports (baseline + spike period) and ask for a CPA diagnosis

Claude Projects setup:

• Download the same repo ZIP

• Upload skill/references/performance_fluctuations.md and skill/references/breakdown_effect.md to your Claude Project

• Upload both CSVs and ask Claude to compare them

Why this repo: It includes Meta's Breakdown Effect documentation — which explains why CPA spikes sometimes have counterintuitive root causes that other tools miss entirely.

4. Audience Overlap Detector

Source repo: irinabuht12-oss/marketing-skills

File: Skills for Claude/Audience Overlap Analysis.md

Claude Projects setup:

• Download the repo ZIP from github.com/irinabuht12-oss/marketing-skills

• Find Skills for Claude/Audience Overlap Analysis.md

• Upload to your Claude Project as knowledge

• Export your ad set data and the audience overlap report from Ads Manager (Ad Sets > select multiple > Inspect > Audience Overlap)

• Upload both to Claude and ask for the analysis

Claude Code setup:

• Copy the file to ~/.claude/skills/audience-overlap/SKILL.md

• Open Claude Code with your data and ask for an overlap analysis

Also available in: AgriciDaniel/claude-ads (part of the /ads audit audience targeting checks)

5. A/B Test Analyzer

Source repo: AgriciDaniel/claude-ads

Included in: The core audit framework (statistical significance is one of the check categories)

Claude Code setup:

• If you already installed claude-ads (see Skill #2), you are set

• If not: curl -fsSL https://raw.githubusercontent.com/AgriciDaniel/claude-ads/main/install.sh | bash

• Drop your test variant CSV into Claude Code and ask for a significance test

Claude Projects setup:

• Download the repo, find ads/references/ folder

• Upload the relevant reference docs to your Project

• Upload your A/B test data and ask Claude to calculate significance

6. Budget Reallocation Modeler

Source repo: mathiaschu/meta-ads-analyzer

Files: skill/references/bid_strategies.md + skill/references/pacing.md

Claude Code setup:

• If you already installed meta-ads-analyzer (see Skill #3), you are set

• If not: download from github.com/mathiaschu/meta-ads-analyzer and copy skill/ to ~/.claude/skills/meta-ads-analyzer/

• Drop your 30-90 day daily breakdown CSV and ask for a budget model

Claude Projects setup:

• Upload bid_strategies.md and pacing.md from the repo to your Project

• Upload your CSV with daily spend data and ask for the analysis

Why this repo: The pacing and bid strategy reference docs teach Claude how Meta's delivery algorithm actually allocates budget — making the reallocation recommendations much more accurate than generic advice.

7. Weekly Performance Report

Source repo: Works with any of the four repos (or none — Claude is good at reporting out of the box). For the best structured output, use AgriciDaniel/claude-ads (run /ads audit for a comprehensive report format) or coreyhaines31/marketingskills (the /paid-ads skill includes reporting frameworks).

Claude Projects setup (simplest):

• No special skill file needed — Claude generates solid reports from CSV data alone

• Upload your weekly CSV export (all levels, all standard metrics)

• Paste this prompt: "Generate a weekly performance report. Include: executive summary (3 bullets), campaign ranking table, top and bottom 3 creatives by ROAS, audience breakdown, and 5 prioritized actions for next week."

Claude Code setup:

• Install either claude-ads or marketingskills (see skills #2 or #8)

• Drop your CSV and ask for a weekly report

8. Ad Copy Generator

Source repo: coreyhaines31/marketingskills

Skill: skills/ad-creative/SKILL.md

Claude Code setup (recommended):

• Open your terminal and run: npx skills add coreyhaines31/marketingskills (this installs all 34 skills at once)

• Open Claude Code and run /ad-creative

• Provide your product brief, audience, and platform specs

Claude Projects setup:

• Download the repo from github.com/coreyhaines31/marketingskills

• Find skills/ad-creative/SKILL.md and upload it to your Claude Project

• Start a conversation and provide your product brief

Also available in: AgriciDaniel/claude-ads (run /ads creative for Meta-specific copy with correct character limits per placement)

9. Competitor Ad Library Analyzer

Source repo: AgriciDaniel/claude-ads

Command: /ads competitor

Claude Code setup:

• Install: curl -fsSL https://raw.githubusercontent.com/AgriciDaniel/claude-ads/main/install.sh | bash

• Go to facebook.com/ads/library and search for your competitor

• Screenshot their active ads or copy the ad text

• Open Claude Code, share the screenshots/text, and run /ads competitor

Claude Projects setup:

• Download the repo, upload ads/references/ docs to your Project

• Screenshot competitor ads from the Meta Ad Library

• Upload screenshots to Claude and ask for a competitive analysis

10. Lookalike Audience Refresher

Source repo: mathiaschu/meta-ads-analyzer

Files: skill/references/ad_auctions.md + skill/references/auction_overlap.md

Claude Code setup:

• If you already installed meta-ads-analyzer (see Skill #3), you are set

• Drop your 90-day ad set CSV (lookalike audiences only, daily breakdown)

• Ask Claude to audit lookalike performance decay

Claude Projects setup:

• Upload ad_auctions.md and auction_overlap.md from the repo to your Project

• Upload your CSV and ask for a lookalike audit

Quick Reference: Which Repo for Which Skill

• 1. Creative Fatigue Scanner — Primary: irinabuht12-oss/marketing-skills | Also in: claude-ads /ads audit

• 2. Wasted Spend Finder — Primary: AgriciDaniel/claude-ads | Command: /ads audit

• 3. CPA Spike Diagnosis — Primary: mathiaschu/meta-ads-analyzer

• 4. Audience Overlap Detector — Primary: irinabuht12-oss/marketing-skills | Also in: claude-ads /ads audit

• 5. A/B Test Analyzer — Primary: AgriciDaniel/claude-ads | Command: /ads audit

• 6. Budget Reallocation Modeler — Primary: mathiaschu/meta-ads-analyzer

• 7. Weekly Performance Report — Primary: Any (or none) | Commands: /ads audit or /paid-ads

• 8. Ad Copy Generator — Primary: coreyhaines31/marketingskills | Command: /ad-creative | Also in: claude-ads /ads creative

• 9. Competitor Ad Library Analyzer — Primary: AgriciDaniel/claude-ads | Command: /ads competitor

• 10. Lookalike Audience Refresher — Primary: mathiaschu/meta-ads-analyzer

Minimum install to cover all 10: Install AgriciDaniel/claude-ads + mathiaschu/meta-ads-analyzer + grab the Creative Fatigue and Audience Overlap files from irinabuht12-oss/marketing-skills. Three repos, full coverage.

Where to Get the Skills (All Free, All Open Source)

Four GitHub repositories contain everything you need. Each has a different strength:

1. coreyhaines31/marketingskills

Best for: Ad creative generation + general paid ads strategy across platforms.

34 marketing skills covering paid ads, ad creative generation, analytics tracking, competitor analysis, and more. The paid-ads skill covers Meta, Google, LinkedIn, and X. The ad-creative skill generates headlines, descriptions, and full ad variations calibrated to each platform's specs.

Works with: Claude Code (slash commands like /paid-ads and /ad-creative).

Link: github.com/coreyhaines31/marketingskills

2. AgriciDaniel/claude-ads

Best for: Comprehensive multi-platform ad audits with weighted scoring.

The most thorough of the four — 186+ checks across 6 ad platforms, with 23 reference documents covering benchmarks, bidding strategies, compliance, conversion tracking, and copy frameworks. The Meta-specific module runs 46 checks including pixel/CAPI validation, creative quality assessment, account structure review, and audience targeting analysis.

Uses parallel agents to run audit and creative tasks simultaneously for faster results.

Works with: Claude Code. Includes an install script that sets everything up automatically.

Link: github.com/AgriciDaniel/claude-ads

3. mathiaschu/meta-ads-analyzer

Best for: Deep Meta-specific analysis — Breakdown Effect, Learning Phase, auction mechanics.

100% focused on Meta Ads with 9 reference documents covering concepts most marketers never learn: why Meta allocates budget to seemingly "worse" audience segments (Breakdown Effect), how the algorithm optimizes during the learning phase and what resets it, auction overlap mechanics, delivery pacing, and performance fluctuation diagnosis.

This is the skill that teaches Claude things even experienced media buyers might not fully understand about Meta's algorithm.

Works with: Claude Code. Optional MCP server available for live API data, but works perfectly fine with CSV exports.

Link: github.com/mathiaschu/meta-ads-analyzer

4. irinabuht12-oss/marketing-skills

Best for: Easiest setup — works with Claude Projects on claude.ai (no terminal needed).

44 marketing skill files covering Meta Ads auditing, creative fatigue detection, audience overlap analysis, retargeting window analysis, frequency cap recommendations, and more. Unlike the other three, these are designed for Claude Projects (the web interface) rather than Claude Code.

If you do not want to install anything and just want to upload files to claude.ai, start here.

Works with: Claude Projects at claude.ai (upload as knowledge files). Also works with Claude Code.

Link: github.com/irinabuht12-oss/marketing-skills

Setup Path 1: Claude Projects (Easiest — No Terminal)

This method works entirely in your browser. No downloads, no terminal, no technical knowledge required.

Time needed: 10 minutes. Cost: Claude Pro subscription ($20/month).

Step 1: Get a Claude Pro Account

Go to claude.ai and sign up. You will need the Pro plan ($20/month) to create Projects with knowledge files. The free plan has limited Project capabilities.

Step 2: Download the Skill Files

Go to github.com/irinabuht12-oss/marketing-skills. Click the green "Code" button, then "Download ZIP." Extract the ZIP file on your computer.

Inside the Skills for Claude/ folder, you will find 44 .md files. The ones most relevant to Meta Ads analysis are:

• Meta Ads Audit

• Creative Fatigue Detection

• Audience Overlap Analysis

• Retargeting Window Analysis

• Frequency Cap Recommendations

• Ad Copy Generator

• Paid Ads Strategy

Step 3: Create a Claude Project

• Go to claude.ai

• Click "Projects" in the left sidebar

• Click "New Project"

• Name it something like "Meta Ads Analyzer"

• In the Project settings, click "Add knowledge"

• Upload the .md skill files you want to use (start with 3-5 most relevant ones)

• Optionally, paste a custom instruction like: "You are a Meta Ads performance analyst. When I share CSV data, analyze it using the frameworks in your knowledge base. Always provide specific, actionable recommendations ranked by impact."

Step 4: Export Your Data from Meta Ads Manager

• Go to business.facebook.com and open Ads Manager

• Select the campaigns you want to analyze (or select all)

• Click the "Columns" dropdown and choose "Customize Columns"

• Add these essential metrics: Performance (Impressions, Reach, Frequency, Results, Cost per Result), Clicks (Link Clicks, CTR, CPC), Conversions (Purchases, Purchase ROAS, Add to Cart, Cost per Purchase), Cost (Amount Spent, CPM), IDs (Campaign Name, Ad Set Name, Ad Name)

• Set your date range (at least 30 days for meaningful analysis, 90 days for trend detection)

• Click the export icon (box with arrow, top right of the data table)

• Choose "CSV" format and click "Export" — the file downloads immediately

Pro tip: For creative fatigue analysis, export with a "Day" breakdown (use the "Breakdown" dropdown and select "By Time" > "Day"). This gives you daily data points for each ad, which is essential for detecting trends.

Step 5: Analyze

• Open your "Meta Ads Analyzer" project on claude.ai

• Upload your CSV file (drag and drop or click the attachment icon)

• Ask your question: "Analyze this data for creative fatigue. Flag any ads where CTR has dropped more than 15% over the last 14 days."

That is it. Claude reads your CSV, applies the analysis frameworks from the skill files, and returns actionable recommendations.

Setup Path 2: Claude Code (More Powerful)

This method uses Claude Code — the desktop/terminal version of Claude that can run skills as slash commands. More powerful, slightly more setup.

Time needed: 15-20 minutes. Cost: Claude Pro or Max subscription ($20-$200/month).

Step 1: Install Claude Code

Download Claude Code from claude.ai/code. Available as a Desktop app for Mac and Windows, a CLI (terminal tool), or a VS Code or JetBrains extension. The desktop app is the easiest for non-technical users.

Step 2: Install the Skills

Option A: One-command install (AgriciDaniel/claude-ads)

Open your terminal (or the Claude Code terminal) and run: curl -fsSL https://raw.githubusercontent.com/AgriciDaniel/claude-ads/main/install.sh | bash

This automatically downloads and installs the skills into the right folder. Once installed, you can use commands like /ads audit and /ads meta.

Option B: Manual install (mathiaschu/meta-ads-analyzer)

• Go to github.com/mathiaschu/meta-ads-analyzer

• Click "Code" > "Download ZIP"

• Extract the ZIP

• Copy the skill/ folder to ~/.claude/skills/meta-ads-analyzer/

• On Mac: open Finder, press Cmd+Shift+G, paste ~/.claude/skills/, create the folder if it does not exist, and drop the files in

• On Windows: navigate to %USERPROFILE%\.claude\skills\ and do the same

Option C: SkillKit install (coreyhaines31/marketingskills)

Run: npx skills add coreyhaines31/marketingskills — this installs all 34 marketing skills at once.

Step 3: Export Your Data

Follow the same Meta Ads Manager export steps described in Setup Path 1, Step 4.

Step 4: Analyze

Open Claude Code and try:

• /ads audit — runs a full account audit against your CSV data

• /ads meta — deep-dive Meta-specific analysis

• /paid-ads — general paid ads strategy and analysis

• /ad-creative — generate ad copy variations

Drop your CSV into the conversation and start asking questions.

Which CSV Export to Use With Each Skill

Not every skill needs the same data. Here is a quick reference:

Creative Fatigue Scanner

• Level: Ad | Breakdown: Day | Key Columns: CTR, Frequency, CPC, Impressions | Date Range: 14-30 days

Wasted Spend Finder

• Level: Ad Set + Ad | Breakdown: None | Key Columns: Amount Spent, Results, Cost/Result | Date Range: 30 days

CPA Spike Diagnosis

• Level: Campaign + Ad Set | Breakdown: Day | Key Columns: All performance metrics | Date Range: Two periods

Audience Overlap Detector

• Level: Ad Set | Breakdown: None | Key Columns: Targeting, Reach, CPM, Frequency | Date Range: 30 days

A/B Test Analyzer

• Level: Ad | Breakdown: Day | Key Columns: Impressions, Clicks, CTR, Results | Date Range: 7-14 days

Budget Reallocation Modeler

• Level: Campaign | Breakdown: Day | Key Columns: Spend, Results, Cost/Result, Reach | Date Range: 30-90 days

Weekly Performance Report

• Level: All levels | Breakdown: None | Key Columns: All performance metrics | Date Range: 7 days

Ad Copy Generator

• No CSV needed — provide your product brief instead

Competitor Ad Library Analyzer

• No CSV needed — use Ad Library screenshots instead

Lookalike Audience Refresher

• Level: Ad Set | Breakdown: Day | Key Columns: CPA, CPM, Reach, Frequency | Date Range: 90 days

"Level" = which tab to export from in Ads Manager (Campaigns, Ad Sets, or Ads). "Breakdown" = whether to add a time breakdown. In Ads Manager, click "Breakdown" > "By Time" > "Day" to get daily rows.

Frequently Asked Questions

Is my data safe?

Your CSV never leaves your machine when using Claude Code. With Claude Projects (claude.ai), your data is sent to Anthropic's servers for processing but is not used for model training on paid plans. No third-party ad analytics service ever sees your data.

Do these connect to my Meta ad account?

No. These skills work with exported CSV files. They do not connect to your account, cannot make changes, and cannot access anything you do not explicitly give them. The one exception is mathiaschu/meta-ads-analyzer, which has an optional MCP server for live data — but the CSV-only mode works perfectly without it.

Can Claude make changes to my campaigns?

Not with these skills. These are analysis-only. To make changes through Claude, you would need to set up MCP with a connector like Pipeboard (covered in our separate guide on Claude MCP for Meta Ads).

Do I need Claude Code or can I use the regular claude.ai?

Either works. Claude Projects at claude.ai is easier to set up (just upload files in your browser). Claude Code is more powerful (slash commands, local file access, no upload step). Start with whichever feels more comfortable.

How much does this cost?

Claude Pro is $20/month. All four skill repositories are free and open source. Meta Ads Manager exports are free. Total cost: $20/month.

What if I also run Google Ads?

AgriciDaniel/claude-ads and coreyhaines31/marketingskills both cover Google Ads alongside Meta. Export your Google Ads data as CSV and use the same workflow.

How often should I run these analyses?

Creative fatigue scans: weekly. Wasted spend audits: bi-weekly. CPA diagnosis: whenever you see a spike. Weekly reports: weekly (obviously). Budget reallocation: monthly or when scaling.