Alex Soltan Facebook ad: “10+ Claude Skills & Agents”

About this ad

Meta Ad Library ID

1695992301400110

Platforms

Facebook, Instagram, Audience Network, Messenger and Threads

Relaunches

1

Niche

AI & Software

How we count

Ad text

Claude built me a $937/day (faceless) low ticket offer at 2.94x ROAS in less than 5 days.⁣ Before I dive into the breakdown itself, I want to give you guys a bit of a disclaimer here:⁣ This post is not for people running 6, 7 or 8 figure businesses.⁣ If that's you, most likely you won't find it valuable or applicable to take your business to the next level.⁣ You're experiencing completely different constraints - most likely solvable by different AI Agents. I can make a post about it later on.⁣ ..This post is for younger me.⁣ It's for people who've been stuck in client work, and cannot find a way to break through that first step of becoming an offer owner…⁣ …creating the offer..⁣ …or for people who've been happily doing client work, but want a more efficient way to build their offers and deliverables for their clients.⁣ If that's you - this post will be a goldmine.⁣ Now when it comes to building an offer using AI, there are usually a few problems people experience..⁣ ..Especially beginners..⁣ Same exact issues I've been dealing with in the early days:⁣ Not having enough time for chatting with AI back and forth to build the products⁣ Experiencing mediocre quality outputs⁣ Not getting consistent outputs - the first page of content looks way too different than the 8th page of content⁣ Very limited context window - after 30 minutes of prompting, you gotta explain back what you're building and working on⁣ Limited integration - most AI tools can't build products within your desired environment, and even if they can, it takes them years to do so, and it eats into your entire context window…⁣ Anyway, there are probably a dozen more, yet those are the most common among people who use AI tools to build a digital product.⁣ Now back to the post..⁣ And I'll start with a back story:⁣ About 4-5 months ago, I had an idea for an offer.⁣ And without thinking too much about it, I started taking action on it.⁣ I started with market research, creating customer avatar profiles, and quickly dove into building the assets.⁣ A few weeks in, I started to struggle big time with my free time / availability to dedicate to this offer.⁣ And for context: the offer was a masterclass series of over 15 videos, plus a bunch of bonuses, templates etc.⁣ Your typical digital product style.⁣ Long story short.⁣ I found myself really struggling to put this offer's deliverables together.⁣ But day by day, I was putting in 5-6 hours of work dedicated just on the deliverables..⁣ And guess what?⁣ I launched the offer, spent $450 in ad spend, and got crickets back.⁣ Now, to say I was upset is to say nothing.⁣ I was fuming (literally)..⁣ But not because the product didn't work out from the first go.⁣ It's because I ignored my own principles and fell in love with my offer, and with the building process, ignoring something more important - validation.⁣ So, I turned off the ads, and rebuilt everything from scratch.⁣ The offer, the ads, the deliverables, the funnel - in under 5 days.⁣ How?⁣ By using Claude AI Agents.⁣ Now, here's the good part.⁣ Instead of going through the offer building process and copy & pasting the outputs manually and formatting everything myself…⁣ I decided to teach Claude my offer building process, gave it access to a few very successful offers I've run in the past, and built not 1, not 2, but 8 AI Agents that work autonomously to build any deliverables I'm telling them to.⁣ Here's a quick breakdown of what those agents look like and what they do:⁣ Btw - if you want me to break down how I used Claude for the marketing part (ads, funnels, sales pages), leave a comment below and I'll cover it in a follow-up post.⁣ A quick note on how these agents actually work:⁣ Each agent has ONE job.⁣ They pass their output to the next agent in line - like an assembly line in a car factory.⁣ Agent 1 decides WHAT to build.⁣ Agent 2 sets up WHERE it lives.⁣ Agent 3 defines HOW it sounds.⁣ Agent 4 structures what each page looks like.⁣ Agent 5 tests before committing.⁣ Agent 6 runs full production.⁣ Agent 7 designs the finished product.⁣ Agent 8 quality-checks everything.⁣ No single agent tries to do it all - because that's exactly where vanilla AI breaks down.⁣ Here's each of them:⁣ - Agent 1 - Product Idea Agent⁣ This is the starting line.⁣ I feed it a rough idea (e.g. "I want to build a Hyrox nutrition playbook for beginner athletes").⁣ The agent turns that one-liner into a locked-in product brief:⁣ Product type (ebook, SOP, workbook, template library, etc.)⁣ Specific niche + target customer avatar⁣ What's FIXED across every page (structure)⁣ What's VARIABLE (the data that changes from page to page)⁣ Why this matters: most people skip this step and end up building a "kinda ebook kinda course kinda swipe file." Mushy products don't sell. This agent forces me to pick a lane.⁣ Output: a clear product brief that every other agent downstream uses as its North Star.⁣ - Agent 2 - Product Build Agent⁣ Once the idea is locked in, this agent spins up the actual Notion database where the product will live.⁣ It creates the database, sets up all the properties (title, status, category, variables), builds the views (Draft, Published, By Category), and generates sample entries to stress-test the structure.⁣ Normally this is a 2-3 hour job of manually clicking around in Notion.⁣ This agent does it in under 90 seconds via the Notion API.⁣ Output: a ready-to-fill Notion database with everything pre-configured.⁣ - Agent 3 - Expert Profile Agent⁣ This is the agent that makes or breaks everything.⁣ Here's the problem: if you let vanilla AI write your content, it'll sound like every other piece of AI slop on the internet. Same vocabulary, same sentence rhythm, same "here are 5 tips for..." energy.⁣ My Expert Profile Agent solves that.⁣ It researches the product domain like a journalist would, then builds a detailed expert persona:⁣ Voice and tone (how would a 20-year veteran in this niche actually talk?)⁣ Vocabulary (what words do they use? what words do they NEVER use?)⁣ Perspective (what's their unique angle on the topic?)⁣ Knowledge boundaries (what they know cold vs. what they don't speak on)⁣ That persona gets injected into every single prompt from this point forward.⁣ Output: a writing persona so specific that readers assume a real human expert wrote every page.⁣ - Agent 4 - Content Blueprint Agent⁣ This agent solves the "page 1 looks nothing like page 8" problem.⁣ It defines the EXACT skeleton every page has to follow:⁣ Which sections appear⁣ The order they appear in⁣ Word counts per section⁣ Formatting rules (bullets vs. paragraphs vs. tables)⁣ Which variables from the database fill which sections⁣ Without this, AI drifts. With this, every page is structurally identical - while the content inside stays unique.⁣ Output: a rigid page template that guarantees consistency across 50, 100, or 500+ pages.⁣ - Agent 5 - Test Content Agent⁣ This is the agent I'm most proud of.⁣ And honestly, the one that's saved me the most money and time.⁣ Why I built it:⁣ On my very first product build, I got excited and skipped straight to full production.⁣ The agent cranked out 80 pages in one shot.⁣ I sat down to review them and… 60 of them were garbage.⁣ Tone was off, the blueprint wasn't being followed tightly enough, and the expert voice was drifting by page 30.⁣ I'd burned through a chunk of credits and had to start over from scratch.⁣ That's when I realized I needed a safety gate BEFORE full production kicks off.⁣ How I came up with the idea:⁣ I stole a concept I've been using in paid ads for years.⁣ You never scale a creative without testing it with $20-$25 first.⁣ Test small → catch problems → THEN scale.⁣ No reason that same principle shouldn't apply to content generation.⁣ How it's built:⁣ The agent sits between the Content Blueprint Agent (Agent 4) and the Generate Content Agent (Agent 6).⁣ Before any full batch run, it pulls 2-3 sample entries from the database and generates only THOSE pages.⁣ Then I review them.⁣ If something's off, the agent traces the problem back to its upstream source:⁣ Bad voice? → The Expert Profile needs tweaking.⁣ Inconsistent structure? → The Blueprint needs tightening.⁣ Weird phrasing? → The prompt needs rewriting.⁣ I fix the upstream artifact, run the test again, and only move to full production once the samples are dialed in.⁣ The result:⁣ Zero wasted batches.⁣ Every full-production run now hits 90%+ quality on the first try, because the real work happens at the sample stage.⁣ This one agent alone has probably saved me 40-60 hours of rework across every product I've built since.⁣ - Agent 6 - Generate Content Agent⁣ Once the samples pass, this agent handles the full batch.⁣ Instead of writing pages one at a time (which is how 99% of people use AI), it spawns multiple subagents in parallel and writes the output directly into Notion.⁣ Example: a 150-page product that would take ChatGPT 10-15 hours of manual prompting gets written in ~90 minutes -dropped straight into the right Notion entries with the right status tags, properties, and formatting.⁣ Output: a fully populated Notion database with every page written, formatted, and tagged as "Published."⁣ - Agent 7 - Design Product Agent⁣ Content alone isn't a product. It has to LOOK like one.⁣ This agent builds the polished homepage customers actually land on:⁣ Callout intro with product overview⁣ Collapsible info toggles ("How to use this," "What's inside," etc.)⁣ Section headers with filtered browse views⁣ Emoji icons on every sub-page⁣ Shareable Notion link ready to drop straight into the delivery email⁣ Output: a finished, professional-looking product I can hand off to a customer without a single manual formatting step.⁣ - Agent 8 - Product QA Agent⁣ I can't manually read 150+ pages to check for quality issues.⁣ But I also can't ship slop.⁣ This agent scans every page against the Expert Profile and Content Blueprint, and flags:⁣ Repetitive phrasing (same hook used 6 times across different pages)⁣ Missing sections⁣ Tone drift (pages that start sounding like generic AI again)⁣ Thin content (sections way under word count)⁣ It outputs a QA report with exactly which pages need regeneration and why.⁣ I approve → it regenerates only the flagged pages → the product is ready to ship.⁣ Output: a polished, consistent product I'd actually pay money for.⁣ Now the part that still blows my mind - actual numbers:⁣ From "I have an idea" to "here's a Notion link I can put behind a paywall" takes 2-5 days now.⁣ And the funnel built around that product is already doing this:⁣ The offer stack:⁣ - Main offer: $37 (currently split-testing a $67 variant)⁣ - Order bump #1: $27 → converting at 63%⁣ - Order bump #2: $37 → converting at 67%⁣ - OTO upsell: $97 → converting at 18%⁣ - Downsell: $47 → converting at 7%⁣ Current performance:⁣ - $200–$300/day in ad spend⁣ - $122 AOV⁣ - 67+ brand new customers acquired⁣ - 180+ total purchases (that's ~2.7 products per customer on average - the bumps and upsells stacking exactly like they're supposed to)⁣ And the part people always ask me about:⁣ I didn't hand-write a single line of the main product.⁣ Every single page was built by the 8 agents above.⁣ Offer: validated.⁣ Deliverable: built entirely by AI.⁣ Scaling: now.⁣ And yes - the ads, funnel, and sales page were built by a completely separate set of Claude agents.⁣ I'll break those down in my next post (comment "ADS" below if you want that one).⁣ If you want to check out the exact 8-agent system I built, clone it, and use it to build your own products… Get the 1-click install by clicking the link below 👇 https://rewyse.ai/the-ultimate-claude-bundle Alex P.S. - I spent nearly 2 years refining this workflow by hand. Now 8 agents do the entire thing in under 5 days. This is the closest thing to "build once, sell forever" that actually feels like cheating.

Where the ad sends people

rewyse.ai

10+ Claude Skills & Agents

Get Instant Access Today!