5 Free Claude Skills That Turn Your Customer Data Into 30+ Winning Ad Angles (Full Setup Guide)

Key Takeaways

• Your next winning ad is already in your data — hiding in reviews, support tickets, competitor ads, and top performers. 5 free Claude skills extract it for you.

• The 5 skills: Review-to-Angle Extractor (pulls phrases from reviews), Competitor Angle Teardown (analyzes their ads), Winning Pattern Decoder (extracts DNA from your top ads), Angle Matrix Generator (30 angles across 6 hook categories), Customer Language Decoder (rewrites in customer voice).

• The skills chain: Skills 1-3 extract what is already working from customers, competitors, and your own winners. Skill 4 expands patterns into 30 angles. Skill 5 rewrites everything in real customer language.

• Everything runs on claude.ai in your browser through Claude Projects. No coding, no API keys, no terminal. Copy-paste the 5 prompts into your project knowledge and you are done with setup.

• Total cost: $65/month (Claude Pro $20 + HeyOz $45). Replaces a $3,000-5,000/month creative retainer with 30+ on-brand ad angles per product.

Introduction

Your next winning Meta ad is already in your data. It is hiding in your reviews, your support tickets, your competitors' ads, and your top-performing campaigns. The problem is not creativity. It is extraction.

Most brands leave this data sitting untouched while their copywriter makes up angles from scratch. The result: generic ad copy that sounds like marketing instead of customers. These 5 Claude skills pull the real angles out of your existing data, expand them into a full matrix of variations, and rewrite everything in your customer's actual voice.

This guide gives you the complete system: how to set up Claude Projects, how to export data from your tools, the full copy-paste prompts for all 5 skills, and how to turn the output into finished ads. No coding. No API keys. Just a Claude Pro subscription and 30 minutes.

What Do These 5 Skills Do?

Skill 1: Review-to-Angle Extractor.

Paste your Shopify, Trustpilot, or any review data. Pulls out the 10 phrases customers use most to describe your product's benefits and pain points. Those phrases become your next 10 ad hooks — in actual customer language.

Skill 2: Competitor Angle Teardown.

Paste competitor ads from the Meta Ad Library. Groups them by hook category, identifies the emotional trigger behind each, and tells you which 5 angles are worth remixing for your brand.

Skill 3: Winning Pattern Decoder.

Paste your top 5 performing ads. Extracts the shared DNA — hook style, format, copy length, CTA structure, emotional trigger. That becomes your template for the next 20 ads.

Skill 4: Angle Matrix Generator.

Product context goes in, 30 distinct angles come out. 6 hook categories with 5 variants each: problem-aware, benefit-led, social proof, direct offer, curiosity, comparison. Each tailored for your product and audience.

Skill 5: Customer Language Decoder.

Drop in DMs, support tickets, and post-purchase surveys. Surfaces the exact words customers use when they describe the problem your product solves. Those words belong in your hook — not your copywriter's vocabulary.

How Do the Skills Chain Together?

Step 1 — Extract what is already working. Skills 1, 2, and 3 pull real data out of three sources: your customers (reviews), your competitors (their ads), and your own performance (your winners).

Step 2 — Expand the patterns. Skill 4 takes the inputs from Skills 1-3 and generates 30 new angles across 6 hook categories. These are not random — they are grounded in what already converts.

Step 3 — Rewrite in customer voice. Skill 5 takes every angle from Skill 4 and rewrites the copy using the exact phrasing customers use in tickets and DMs. Your ads stop sounding like AI-generated marketing and start sounding like the conversations your buyers are already having.

The output: 30+ ad angles per product, all grounded in real data, all ready to produce.

What Do You Need Before Starting?

Claude Pro ($20/month) at claude.ai — the free plan has message limits that make this workflow hard to run. A browser — everything runs on claude.ai through Claude Projects. No terminal, no installations.

Data sources (gather what you have): product reviews (Shopify, Trustpilot, Google, Amazon), support tickets (Gorgias, Zendesk, Intercom, email), competitor ads from the Meta Ad Library (screenshots or exports), performance data on your top 5 ads, and any DMs or post-purchase surveys.

For production: HeyOz ($44.99/month at heyoz.com) to turn angles into finished ads. Or use Canva, Figma, or any design tool. Total cost: $65/month for the full ideation + production stack.

How Do You Set Up Your Claude Project?

Claude Projects lets you give Claude persistent context — files and instructions it remembers across every conversation in that project.

Go to claude.ai, click Projects in the left sidebar, click New Project, name it Ad Angle Ideation System, and click Create.

Click Set project instructions and paste: You are a Meta ad creative strategist. You have 5 specialized skills loaded as knowledge (Review-to-Angle Extractor, Competitor Angle Teardown, Winning Pattern Decoder, Angle Matrix Generator, Customer Language Decoder). When I ask you to run a skill, follow its exact output format. When I ask you to chain skills, use the output of the previous skill as input. All outputs must be Meta-spec compliant: headlines under 40 characters, primary text under 125 characters first line, descriptions under 30 characters, no ALL CAPS, no excessive claims. Match my brand voice once established. Prioritize customer's actual words over marketing speak.

Save each of the 5 skill prompts (in the next section) as separate knowledge items. Click the + in the knowledge section, select Add text, paste the skill prompt, save, and repeat for all 5. Claude now has access to all skills in every chat.

What Are the 5 Skill Prompts?

All 5 prompts below are copy-paste ready. Save each as a separate text entry in your Claude Project knowledge (click the + in the knowledge section, select Add text, paste the prompt, save).

Skill 1 — Review-to-Angle Extractor

Save as: Skill 1 — Review to Angle Extractor

WHEN TO USE: User pastes customer reviews and asks for ad angles or customer language extraction. PROCESS: Read all reviews provided. Separate 5-star/positive from 1-3 star/negative reviews. Extract specific phrases (not paraphrased summaries). Group by theme. Identify the 10 strongest phrases across all reviews.

OUTPUT FORMAT — Customer Benefits (from positive reviews): Extract the top 5 phrases where customers describe OUTCOMES they got from the product. Use their EXACT words. For each phrase: the exact quote (word-for-word from a review), why this works as an ad hook (1 sentence), suggested headline using this phrase (under 40 chars), suggested primary text opening line (under 125 chars).

OUTPUT FORMAT — Customer Pain Points (from both positive and negative reviews): Extract the top 5 phrases where customers describe the PROBLEM they had BEFORE using the product, or unresolved frustrations. For each phrase: exact quote, why this pain resonates (1 sentence), suggested headline using this pain point (under 40 chars), suggested primary text opening line (under 125 chars).

OUTPUT FORMAT — Trigger Events: From the reviews, identify any mentions of WHEN or WHY people started looking for this product. What was happening in their life? List up to 5 trigger events.

OUTPUT FORMAT — Vocabulary Fingerprint: List 15-20 specific words or phrases that appear repeatedly across reviews. These become your brand's customer vocabulary. Words to avoid (too generic): good, great, amazing, love it. Words to keep: specific, concrete, sensory, emotional language. RULES: Never paraphrase customer quotes — use exact words. Include reviewer context if available (age, gender, use case). Flag any review that contains a trigger event, objection overcome, or comparison to a competitor — these are gold.

Skill 2 — Competitor Angle Teardown

Save as: Skill 2 — Competitor Angle Teardown

WHEN TO USE: User pastes competitor ads (copy, screenshots, Ad Library exports) and asks for analysis. PROCESS: Identify each unique ad, categorize by hook type, score by likely performance signals, identify remix opportunities. HOOK CATEGORIES (group each ad into one): 1. Problem-Aware (leads with pain point). 2. Benefit-Led (leads with outcome). 3. Social Proof (leads with numbers/reviews/authority). 4. Direct Offer (leads with price or deal). 5. Curiosity (creates information gap). 6. Comparison (positions against alternatives).

OUTPUT FORMAT — Ad Teardown Table: For each ad, columns for Competitor, Hook Text, Category, Emotional Trigger, Format, Estimated Performance Tier. Performance tier inference (when data is limited): Tier 1 (likely winner) = running 30+ days OR 5+ variants OR in 5+ countries OR appears in 3+ formats. Tier 2 (working) = 2+ variants OR running 14+ days. Tier 3 (testing) = new, single variant, single format.

OUTPUT FORMAT — The 5 Angles Worth Remixing: Pick the 5 most promising angles for the user to adapt. For each: why this angle works (psychological principle), what to KEEP when remixing (structure, emotion, format), what to CHANGE (claims, specifics, brand references), a sample remix headline for the user's product, which hook category it fits.

OUTPUT FORMAT — Gaps: List 3-5 hook categories or emotional angles that competitors are NOT covering heavily. These are whitespace opportunities. RULES: Never suggest copying exact text or creative assets. Focus on the psychological trigger, not the specific words. If competitor data is thin, say so — don't fabricate signals.

Skill 3 — Winning Pattern Decoder

Save as: Skill 3 — Winning Pattern Decoder

WHEN TO USE: User pastes their top-performing ads (with performance data when available) and asks for pattern analysis. PROCESS: Analyze all winning ads together, looking for shared elements. Identify the DNA — the common structural and psychological patterns across winners. Produce a reusable template.

OUTPUT FORMAT — Shared DNA Analysis. Hook Structure: What opening formula appears across winners (question, bold claim, statistic, story, command). Average hook length in words/characters. Sentence structure (short/punchy, compound, narrative). Emotional Triggers: Which 2-3 emotions dominate (fear, aspiration, curiosity, validation, urgency, relief, belonging, FOMO). How are these triggers delivered (directly, implied, through story, through contrast).

OUTPUT FORMAT — Copy Formula: How is the primary text structured (Problem-Agitate-Solve, Benefit-Proof-CTA, Story-Lesson-Offer, Feature-Advantage-Benefit, List of Results, Testimonial-led). Average length in words. Short-form/medium/long-form preference. Format Patterns: Static vs video vs carousel distribution. If video, hook type in first 3 seconds. If static, image style (product-focused, lifestyle, comparison). CTA Patterns: Which CTAs appeared, placement (standard button vs in-copy), action verbs used.

OUTPUT FORMAT — The Winning Template: Based on the DNA, produce a reusable template with placeholders formatted as Hook type, specific format, Body structure with X elements, CTA style. The Do-Not-Do List: What patterns should be AVOIDED based on what winners have in common and what they do NOT do (e.g., Winners never use all caps, Winners avoid questions longer than 8 words). Next 20 Variations Brief: Give a 1-paragraph brief the user can feed to Skill 4 to produce 20 new variations that follow this template. RULES: Look for patterns across at least 3 of the 5 winners before declaring something part of the DNA. When performance data is available, weight more heavily toward highest-ROAS winners. If winners are too different to find a pattern, say so and suggest more data.

Skill 4 — Angle Matrix Generator

Save as: Skill 4 — Angle Matrix Generator

WHEN TO USE: User provides product context and wants a full matrix of ad angles. REQUIRED INPUTS: Product name and URL, price point, 3-5 key benefits, target audience description, main competitors, brand voice notes (if available), any winning patterns from Skills 1-3 (if available).

OUTPUT FORMAT: Generate 30 angles across 6 hook categories (5 per category). Each angle must be distinct — no two should convey the same idea with different words. Category 1 Problem-Aware (5 angles): Lead with the pain point, best for cold audiences. Category 2 Benefit-Led (5 angles): Lead with outcome or transformation. Category 3 Social Proof (5 angles): Lead with numbers, credibility, reviews. Category 4 Direct Offer (5 angles): Lead with price, discount, deal, best for warm audiences. Category 5 Curiosity (5 angles): Create information gap, drives high CTR but needs strong landing page. Category 6 Comparison (5 angles): Position against alternatives or the old way.

OUTPUT FORMAT — For each of the 30 angles provide: Headline (under 40 chars), Primary text first line (under 125 chars), Primary text body (up to 300 chars), Suggested CTA (Shop Now, Learn More, Get Started), Emotional trigger (core emotion this taps), Audience segment this works best for, Format recommendation (static, video, carousel). Scoring: Rate each 1-10 on hook strength (would it stop a scroll), brand fit (does it match the product and audience), Meta compliance (character limits, no restricted claims). Rank all 30 by average score. Flag the top 10 for production.

RULES: Use customer language from Skill 1 outputs if provided. Follow the winning template from Skill 3 outputs if provided. Avoid duplicates — each angle should target a different trigger, audience segment, or format. Meta-spec check: no ALL CAPS, no excessive emoji (max 2), no before/after claims without disclaimers, no health claims implying results without proof.

Skill 5 — Customer Language Decoder

Save as: Skill 5 — Customer Language Decoder

WHEN TO USE: User pastes DMs, support tickets, post-purchase surveys, or any direct customer communication and asks for language extraction. PROCESS (based on the Sales Safari framework): Analyze the data through 4 lenses. 1. Pain — stated problems and frustrations. 2. Jargon — how customers speak, their specific vocabulary. 3. Worldview — their assumptions, priorities, what they value. 4. Recommendations — how they frame solutions, what they compare to.

OUTPUT FORMAT — Pain Inventory: List 10-15 specific pains. For each: exact customer phrase (verbatim), frequency (how many customers mentioned it), severity (1-5 mild annoyance to dealbreaker), whether the product already solves it, suggested ad hook if solved. Jargon Dictionary: List 20-30 specific words or phrases customers use that are NOT standard marketing language — gold for ad copy. For each: the word/phrase, the context it appears in, whether positive/negative/neutral. Example good entries: night and day difference, finally clicked, the only thing that worked. Bad entries (too generic): good, love it, recommend.

OUTPUT FORMAT — Worldview Insights: What do customers believe? What do they think the category/industry gets wrong? Who do they trust? Who do they distrust? What are they afraid of? What does success look like to them? Switching Language: If the data contains any mentions of switching from alternatives, extract what they tried before (by name if possible), why it did not work, the moment they decided to switch. These are powerful Comparison category ad hooks.

OUTPUT FORMAT — Objection Database: List every objection, hesitation, or concern expressed. For each: the objection (verbatim), what the customer needed to hear to overcome it, the segment that raises this objection most. Ad Copy Rewrite Kit: Take 5 generic ad copy samples from the user (or generate 5 from their product) and rewrite each using pure customer language. Show before/after. Example — Before: Our serum delivers noticeable results in just 2 weeks. After: The weird thing is, I noticed the difference in like 12 days. I wasn't even looking for it.

RULES: Pull DIRECTLY from the source data — do not invent phrases. Flag any phrase that appears in 3+ separate pieces of data as signal (high-frequency = likely to resonate). Note when a piece of customer language contradicts the brand's current marketing — these are the biggest opportunities.

How Do You Export Your Data?

Shopify Reviews:

Shopify discontinued its native Product Reviews app. Most stores use Judge.me (Admin > Judge.me app > Export Reviews > CSV), Yotpo (Dashboard > Analytics > Export), Loox (Dashboard > Reviews > Export), or Stamped.io (Reviews tab > Export). Export at least 50-100 reviews from the last 3-6 months for best Skill 1 output.

Trustpilot Reviews:

On paid plans: Dashboard > Reviews > Export (admin role required). For free accounts: use Reviewflowz or Outscraper for bulk extraction.

Support Tickets:

Gorgias: Tickets > Filter (last 90 days) > Export CSV. Zendesk: Reports > Export > Ticket data. Intercom: Settings > Data > Export. Export at least 100 tickets, skip one-line help tickets.

Post-Purchase Surveys:

KnoCommerce, PostPilot, Fairing export CSV responses. Google Forms: Responses tab > Download CSV.

Meta Ad Library (competitor ads):

Manual: go to facebook.com/ads/library, search your competitor, filter by Active ads, screenshot 10-20 ads, paste into Claude. Automated: use TryPeggy's Facebook Ads Library MCP (github.com/trypeggy/facebook-ads-library-mcp).

Your Top 5 Ads:

Meta Ads Manager > filter by last 30-60 days > sort by ROAS or CTR > copy top 5 with headline, primary text, description, CTA, and performance metrics.

How Do You Run the Full Pipeline?

Session 1 — Extract What Is Working (Skills 1, 2, 3):

Start a new chat. Run Skill 1 by saying: Run Skill 1 — Review-to-Angle Extractor on the following reviews, then paste 50-100 reviews. Save the output. Run Skill 2: Run Skill 2 — Competitor Angle Teardown on these competitor ads, then paste details for 3 competitors. Save the output. Run Skill 3: Run Skill 3 — Winning Pattern Decoder on my top 5 performing ads, then paste headline, primary text, ROAS, CTR for each. Save the output.

Session 2 — Generate the Angle Matrix (Skill 4):

Start a new chat in the same project. Paste: Run Skill 4 — Angle Matrix Generator. Product [name, URL, price, benefits, audience, competitors, brand voice]. Incorporate these inputs from my earlier runs: [paste Skill 1 output, Skill 2 output, Skill 3 output]. Generate 30 angles across 6 hook categories using these inputs. You get 30 angles grounded in customer reviews, competitor gaps, and proven patterns.

Session 3 — Rewrite in Customer Voice (Skill 5):

Paste: Run Skill 5 — Customer Language Decoder on this data, then paste 50-100 support tickets, any DMs, and survey responses. After extracting the pain inventory, jargon dictionary, and objection database, rewrite the top 10 angles from Skill 4 using pure customer language. You end up with 10 polished, customer-voice-approved angles ready for production.

How Do You Turn Angles Into Finished Ads?

Using HeyOz (fastest): go to heyoz.com, add your product (paste URL to auto-import images and brand colors). For each angle: select format (static, UGC video, carousel, product demo), paste the headline, paste the primary text, choose product image matching the angle, generate. Export at Meta specs: 1080x1080 (feed), 1080x1350 (4:5), 1080x1920 (Stories/Reels).

Using Canva or Figma: create a brand template, duplicate for each angle, swap in copy and images, export. Using AI image generators: paste visual direction from Skill 4 output into Midjourney, DALL-E, or Nano Banana. Add text overlays in Canva.

Testing framework: one campaign, one ad set targeting core audience, 10 creatives as individual ads at $5-10 per ad per day for 3-4 days. Kill below 1% CTR with zero conversions, scale top 3-5. Feed winners back into Skill 3 for next cycle.

What Does This Cost?

Claude Pro: $20/month. All 5 skill prompts: free (included in this guide). HeyOz Basic (production): $44.99/month. Meta Ads Manager: free. Total: $65/month.

Compare to: a creative agency retainer at $3,000-5,000/month, a freelance copywriter at $500-2,000 per round, a standalone ad intelligence tool (Foreplay) at $249/month, or a voice-of-customer research agency at $5,000-15,000 per project. This system gives you ideation, pattern extraction, language mining, and production for $65/month.

Frequently Asked Questions

Do I need to know how to code?

No. Everything runs on claude.ai in your browser. The only technical step is copy-pasting the skill prompts into your project knowledge.

What if I do not have enough reviews yet?

Use competitor reviews. Their Trustpilot pages and Amazon listings contain your target customer's language too. Mine them with Skill 1.

How long does the full pipeline take?

First time: 2-3 hours while you gather data and learn the sequence. Once you know the workflow: 45-60 minutes per product cycle.

Will all 30 angles from Skill 4 be good?

Expect 15-20 strong, 10 average, a few misses. The scoring step inside Skill 4 flags the top 10. Skill 5 polishes those into finished copy.

Can I use this for Google Ads, TikTok, or LinkedIn?

Yes. Hook categories and customer language extraction work for any platform. Adjust character limits for the target platform (Google RSA: 30-char headlines, 90-char descriptions; TikTok: captions under 100 chars).

How is this different from just asking ChatGPT to write ad copy?

Three differences. The skills are structured frameworks that enforce specific outputs. They chain together so each output informs the next. They are grounded in YOUR data (reviews, tickets, competitor ads, winners) instead of AI guessing what your brand should sound like.

What if my brand has no winning ads yet?

Skip Skill 3 for now. Run Skills 1, 2, 4, and 5 without the winning pattern input. Once you have 3-4 weeks of real performance data, re-run the full pipeline with Skill 3 included.

How do I know the customer language is converting?

Track it in Meta Ads Manager. Launch 10 ads with customer-voice copy and 10 with polished marketing copy — the customer-voice ads typically outperform on CTR and cost per click. Let the data decide.