INTRO
Hey there DTC operators, welcome to this week’s edition of DTC+AI Unhyped.
In today’s edition, we’re covering the most underrated Claude feature that will save you hours of time a week (literally), and how to use it. At the very end, I’m also sharing an ecommerce email revenue forecasting agent that you can use to “wow” your boss (or clients).
Let’s dive in.

Claude Skills: The AI Feature 95% of eCom Marketers Are Sleeping On
I put 10+ hours into learning Claude Skills (and counting), but it was worth it. It’s turned multi-hour projects into 5-minute tasks across a bunch of different areas.
For instance, I’ve now created Claude Skills for personal use, like workout plans, meal planning, financial planning, and how to win arguments with my wife (kidding, no skill helps me there).
I’ve also created Claude Skills for professional use, like revenue forecasting, personnel planning, sales coaching, email campaign creation, ad creation, and the likes.
In fact, it is so fun and easy to do now, I literally just spin up Claude Skills for friends of mine. The same way I’ve vibe coded a few basic apps for some agency friends over the last two weeks (keeps me sharp on new tools and helps a friend out).
What are Claude Skills?
Claude Skills aren’t that new, so I wasn’t planning on writing anything about it. But every operator or agency owner save for one that I’ve spoken to has no idea about Claude Skills… so it’s time.
Skills are basically sub-agents within Claude’s web and desktop app. That’s the fancy, AI-way of saying it’s like creating a specialized AI employee. No prompts. No templates. Just sub-agents that know how to do specific jobs based on how you told them to.
Wait, WTF is an Agent Again? (Quick Detour)
Before we get into Skills, let’s clarify what an AI agent actually is, because the term gets thrown around like confetti at a wedding.
Basic AI (Most interactions with ChatGPT, Gemini):
You ask a question
It gives an answer
Done
“AI Mode” in Google is a great example of this.
AI Agent:
You give it a goal or task
It figures out what steps to take
It can use tools (search web, analyze files, run code)
It makes decisions about what to do next
It completes the task
A simple way to think of it is like the difference between asking someone for directions vs. asking them to drive you there vs. asking them to drive you there, stop for coffee, pick up your dry cleaning, and remember that you hate the highway.
Basic AI is the first one. Agents are supposed to be the third one. But most agents right now are that friend who agrees to drive you but then gets lost, calls you six times for directions, and somehow ends up at a Wendy’s drive-thru asking if you really meant ‘Target’ or if maybe you actually wanted ‘Jimmy Johns’ (I’m locked into this bread theme for this newsletter now).
Examples of agents:
ChatGPT’s Agent trying (and failing) to order my groceries
Perplexity’s Comet browser that actually shops for you
Using Sub-agents in Claude Code for activities like performance marketing
Raleon’s Email Creator Agent
Back to Why You Should Care About Claude Skills
Skills are specialized sub-agents that live inside Claude. They’re not just saved prompts (though that’s part of it). They’re mini AI specialists with:
Specific knowledge - They understand your business, your processes, your data
Defined workflows - They know the exact steps to complete specific tasks
Tool access - They can search your past chats, reference your project files, analyze data
Consistent behavior - They approach tasks the same way every time
Think of Skills like hiring specialists vs. generalists:
Claude without Skills: You’re talking to a smart generalist who can help with anything but needs full context every time.
Claude with Skills: You’re talking to a specialist who already knows your business and has the exact workflow down.
What makes this even more powerful is Skills can reference your past conversations and project files either by you asking when you tee up the prompt, or because you built it directly into the Skill.
So instead of:
Hey ChatGPT, here’s my email data [paste], here’s my business context [paste], here’s what I need [explain]…You do:
Look at my past chats about email ideas and use my project [project name] with my historic performance data to forecast next week’s campaign revenue.Or even better
Forecast next week’s campaign revenueAnd it automatically looks up past chats for ideas, and your project, because when you created the Skill, you had Claude build that into it for you (we’ll get to building Skills in just a second. It’s so easy).
How to Create Claude Skills
Now that you know Claude Skills are the most handy thing since sliced bread, let’s get into how to create your own.
Just as an aside - sliced bread is actually super handy. Have you tried cutting a loaf yourself? Those slices are inconsistent and terrible. Just like trying to get ChatGPT to maintain consistent brand voice across three different conversations.
It’s these sorts of brilliant asides that help you know this newsletter is not AI-generated. An AI would have said something boring like ‘sliced bread revolutionized food convenience in 1928.’
On to creating Skills.
How to Enable Skills (Takes 2 Minutes)
First up, you have to turn on Claude Skills.
Step 1: Log in to your Claude.ai account and go to Settings
Step 2: Head to Capabilities → Scroll to the bottom → Toggle “Skills” on
Step 3: Toggle the “Skill Builder” skill on. This is an important one as it lets Claude build skills for you.
Step 4: To add a new skill, like the at the end of this newsletter, you’ll upload it here
That’s it! You’re ready to start using and creating skills.
Building Your First Claude Skill
This is my favorite part, because it’s just so easy. Let’s say you wanted to build a Skill around forecasting email revenue (I have a version of this for you at the end).
You would prompt Claude with something like:
Create a new skill that forecasts email revenue for my ecommerce brand.
I want you to use the historical data I provide to create the estimated forecast.
You should forecast both my automations revenue and my campaign revenue based on past performance.
When forecasting, come up with 3 scenarios: Aggressive, normal, and conservative.
Once you’ve come up with forecast, create them as charts in a react app I can view in Claude that’s interactable so I can share it with my CEOThat last part about charts is a little pro tip for how to save yourself extra time so you’re not messing around with excel files.
That’s it! Once you tee that up, you’ll see Claude spend time working on your new skill. It should also give you instructions on how to use it. When it’s done, you should be given a download link by Claude. If you don’t get a download link, ask it for one. Sometimes Claude gets lazy.
From there, download the skill, and upload it to your Claude account following the steps provided earlier. If Claude didn’t create the skill, make sure you go back and enable the “Skill Creator” skill.
It really is that easy to ask Claude to create new skills for you. It’s honestly what makes it kind of addicting. You just ask it to make the skill and experiment with it until it’s how you want it.
Here’s a few other ideas to play around with that I’ve found very helpful:
Competitive analysis
Checking competitor prices
Email campaign ideas
Landing page optimization
Customer review synthesis
Ad copy angles
A Free Email Revenue Forecasting Skill for You
For those who survived it past all my unexpected bread comments, you can download and enjoy my email revenue forecasting skill.
Once you’ve uploaded it to your Claude Skills, using it is easy.
Use the historical data from [project name] project and run my revenue forecast skill for January.If you don’t have a project, you can just copy and paste the data in.
The OTHER Reason You Should Care About Skills
Here’s my prediction: In 12 months, using AI without specialized agents (Skills or whatever other platforms call them) will feel as outdated as manually tracking inventory in a spreadsheet.
The brands winning with AI aren’t the ones using ChatGPT for one-off tasks or generic work. They’re the ones who have built their own systems and implemented others from software. Specialized agents compound. Your data compounds. That means your workflows get more efficient every time.
You don’t need to build 47 Skills tomorrow. Start with one. Pick your most repetitive, context-heavy task. Build a Skill for it. Use it for a month. Share it with your team.
Then build the next one.
Start to store all your brand knowledge and use skills, and in 3 months, you’ll have a library of specialized AI agents that know your business better than most agencies. And because they can reference past chats and project files, they’ll get smarter over time as you feed them more context.
And you’ll wonder how you ever kept copy-pasting context into ChatGPT every morning instead of saving 6 hours a week and using that time to do literally anything else. Like slice some bread.
This Week’s Rabbit Holes
Sam Altman’s “Code Red” Memo - Regular readers won’t be surprised by this. I’ve been calling OpenAI’s struggle for months now.
ChatGPT Image 1.5 Launch - I’ll cover this beginning of the new year, but you should 100% check out 1.5 Image. Nano Banana has been my go to, but there’s some good stuff here.
Claude Skill Documentation - If you want to go even deeper than what I covered here
Klaviyo shared their BFCM numbers - Biggest highlight to me is owned channels are still massive. Both on SMS and email
And that’s it for this week’s edition. Stop explaining your business to LLMs every day. Build Skills. Compound your AI workflows.
Your future self (who just saved 6 hours this month) will thank you.
No newsletter next week, but I wish everyone a Merry Christmas. I’ve got some great AI-breakdowns coming in the new year!
Have you been liking these more in-depth emails, or did you prefer a little higher level from before? I’ve been loving all the emails you all have been sending, so let me know!
