Running Meta ads manually is a full-time job. Every week you are manually scrolling the Ad Library to see what competitors are running. You are briefing copywriters and waiting days for variations. You are spending money on creative before you know if the hook, copy, or CTA is any good. Agencies charge $5-10K/month to handle this. Claude does it in minutes.
These are 5 Claude Code skills that replace the manual work of running a Meta ads operation.
Skill 1 — /spy
What it does: Pulls every active ad from any Facebook page via the Meta Ad Library API. Diffs against last week's pull to surface only new creative. Returns a structured intelligence report with hooks, CTAs, offer types, creative angles, and how long each ad has been running.
Which hooks your competitors are testing right now. Which ads have been running 30+ days (the winners). What new creative they launched this week. Prerequisite: You will need a Meta Ad Library API access token set as $META_ACCESS_TOKEN in your environment.
Skill 2 — /competitive-ads-extractor
What it does: Give it 3-5 competitor names or URLs. It pulls their active ads, ranks hooks by how often they appear, identifies offer structures and CTA patterns, then outputs a gap analysis showing emotional angles your competitors are not using that you could own.
Top hooks ranked by frequency. Offer structure patterns across competitors. CTA analysis. Gap analysis — angles nobody in your market is running.
Skill 3 — /bulk-creative
What it does: Give it your product description and your CLAUDE.md brand context. Generates 20 distinct ad copy variations — short, medium, and long format — in about 10 minutes. What used to take a copywriter 4-8 hours. Make sure your CLAUDE.md has your brand voice, your target audience, and your core offer before running this.
Skill 4 — /ads meta
What it does: Runs 186 checks across your Meta account. Detects creative fatigue, audience overlap, CPM anomalies, and frequency violations. Returns a unified health score from 0 to 100 with a prioritised action plan. Creative fatigue before your ROAS drops. Audience overlap eating your own budget. Frequency violations and CPM anomalies. Placement underperformance.
Skill 5 — /ads-score
What it does: Paste in any ad — yours or a competitor's. It scores it across 6 dimensions on a 1-10 scale and tells you exactly where it is weak and why — before you put money behind it. Hook strength. Copy effectiveness. CTA clarity. Emotional resonance. Offer structure. Visual-copy alignment. Run this on every ad before launch.
How to Chain Them
Run /spy on Monday morning to see what competitors launched last week. Run /competitive-ads-extractor on your top 3-5 competitors to find the gaps. Use /bulk-creative to generate 20 variations targeting those gaps. Run /ads-score on your top 5 variations before launch. Run /ads meta weekly to catch fatigue and overlap before they hurt your numbers. The whole workflow takes under an hour. Agencies charge a monthly retainer for this.
Setup Checklist
Install Claude Code: npm install -g @anthropic-ai/claude-code. Authenticate: run claude and sign in. Create your CLAUDE.md with your brand voice and audience. Set your Meta API token: export META_ACCESS_TOKEN=your_token_here. Install the skills from the Composio skills repository.