Founders, media buyers and marketers at e-commerce brands get 12 Claude agents for Meta Ads: one job each, one /meta command, and a person who approves every change.
<aside> 🚀
🚀 FREE STRATEGY CALL Want to grow your product through viral reach? We help teams across iGaming, mobile apps, AI products, and SaaS build creator-led organic distribution generating tens of millions of views at CPMs starting from $0.03.
<aside> 🚀
</aside>
On the call you'll receive:
| In the post | Count | Where it lives on this page |
|---|---|---|
| Claude agents for Meta Ads | 12 | The 12 agents section |
| Type /meta and pick one | 1 | Set up /meta in Claude Code |
| Account Check: finds the ad sets wasting spend | 1 | Toggle 01 in The 12 agents |
| Pixel Check: matches Meta purchases to store orders | 1 | Toggle 02 in The 12 agents |
| Angle Map: who buys and why, from your reviews | 1 | Toggle 03 in The 12 agents |
| Hook Lab: 30 openers for every angle | 1 | Toggle 04 in The 12 agents |
| Creator Brief: a one-page brief per concept | 1 | Toggle 05 in The 12 agents |
| Static Builder: turns reviews into static ads | 1 | Toggle 06 in The 12 agents |
| Variant Maker: 1 winning ad into 15 versions | 1 | Toggle 07 in The 12 agents |
| Launch Check: names, budgets and links before go-live | 1 | Toggle 08 in The 12 agents |
| Test Reader: calls winners on day 3 and day 7 | 1 | Toggle 09 in The 12 agents |
| Fatigue Watch: flags ads losing their hook | 1 | Toggle 10 in The 12 agents |
| Budget Mover: shifts spend to winners in steps | 1 | Toggle 11 in The 12 agents |
| Monday Report: spend, MER and next week's tests | 1 | Toggle 12 in The 12 agents |
| Openers for every angle | 30 | Hook Lab output. All 30 are in the prompt pack |
| Versions from 1 winning ad | 15 | Variant Maker output. All 15 are in the prompt pack |
| Reads on day 3 and day 7 | 2 | Test Reader, two worked examples |
| Copy ready prompts | 13 | One in each agent toggle, plus the handover note |
Also here: The folder layout and the /meta command file. The brand context file. The Claude Projects version. The handoff diagram. The Fernway run, end to end. Five imperfect inputs, and who decides.
Jump to: Start here. Set up /meta in Claude Code. How the agents hand off. The 12 agents. The Fernway run, end to end. How to use it every week. How this was made.
brand-context.md with your own.Five steps. Under twenty minutes. One real read of your own account.
<aside>
How to run any prompt on this page. Open a Claude chat. Paste the prompt text. Paste your brand context and your inputs below the last line. In Claude Code, type /meta and the agent word instead. Set up /meta in Claude Code shows how. Without Claude Code, use the Claude Projects version in the same section.
These agents are not a running service. Each one runs when you call it. Claude reads, compares and drafts from the exports you add. People launch, spend, message and decide. No agent changes a live campaign. Nothing goes live without your OK. Views are not purchases. Before using these prompts you should omit any PII (personal identifiable information) and check with your compliance team and/or your AI policy.
</aside>
<aside>
Downloads. The prompt pack holds all 13 prompts, the sample inputs and every output in full, with all 30 openers and all 15 versions. The test log holds 12 sample rows and a header row. The zip holds both files plus the meta-agents Claude Code project. It has the /meta command, 12 agent files, the brand context file, a test log copy, an exports note and the Claude Projects instructions. That is 19 files in all. The .claude folder inside meta-agents starts with a dot, so some file browsers hide it. On a Mac, press Cmd, Shift and full stop in Finder to show it.
12-claude-agents-meta-ads-prompts.md
12-claude-agents-meta-ads-pack.zip
</aside>
The idea in one line: One job each. Type /meta, pick the agent, approve the change. Each agent is a file with one job. One command calls any of them. One brand context file feeds all twelve, so no agent starts from a blank chat.
Checked against Claude Code documentation, 21 November 2026. Project subagents live in .claude/agents/ as Markdown files. Each starts with a name and a description. Skills live in .claude/skills/, one folder each, with a SKILL.md file. The folder name becomes the command, so the meta folder gives /meta. Older .claude/commands/ files still work. Words typed after the command reach the skill as $ARGUMENTS. A subagent is called by naming it in a request.