Last time, we sorted out what those mysterious little markdown files are (you can catch up here: What Are Markdown Files?). Right at the end, I mentioned something called Skills and promised they deserved a post all of their own. Well, here we are.
If you’ve been hearing the word “Skills” floating around your AI tools lately and quietly nodding along while having no real idea what people mean, you’re in exactly the right place. I get this question a lot from clever business owners who are using AI every day but keep bumping into new bits of jargon. So let’s clear this one up properly. Plain English, no tech headache, and a genuinely good reason you’ll want to start paying attention to Skills.
So, what is a skill, really?
A Skill is a saved set of instructions that teaches your AI how to do a specific task properly, the same way, every single time.
That’s the heart of it. Instead of explaining the same task to your AI over and over (and getting a slightly different result each time because you worded it slightly differently), you write the instructions down once, save them as a Skill, and your AI follows them whenever that task comes up.
Think of it like a really good recipe card you hand to a brilliant but literal assistant. The recipe spells out the steps, the ingredients, the order, and the little details that make it turn out right. Give that same card to your assistant on a Monday or a Friday, when you’re fresh or frazzled, and you get the same lovely result either way. That’s a Skill.
Most Skills are written in Markdown (yep, the plain little file type from last time), saved in a file called SKILL.md. It holds a name, a short description of when to use it, and the step-by-step instructions for getting the job done.
A prompt, a template and a Skill walk into a bar
These three get muddled all the time, so let’s draw a clear line between them. Once you see the difference, the whole thing clicks.
A prompt is a one-off instruction. You type it, you get a result, and then it’s gone. Next time you want the same thing, you’re typing it all again from scratch (or digging through your chat history trying to find it). If you’re organised, you may save it somewhere for easy access and create a prompt template that you can copy and paste, maybe replacing [X] with the appropriate variable. But from my experience, most people don’t.
A Custom GPT or Gem is a whole assistant you set up once for a particular job, then go back to whenever you need it. (Custom GPTs live in ChatGPT, Gems live in Gemini, and they’re the same idea.) This is a big step-up because the instructions and knowledge are baked in, so you’re not starting from scratch. The catch is that it’s a separate destination. You have to stop what you’re doing, switch into that specific assistant, and work inside it, and it doesn’t retain any knowledge from outside that individual chat. Annoying!
A Skill flips that around. Instead of you going off to a separate assistant, the Skill comes to you. It lives inside the AI tool you’re already using, and it switches itself on automatically the moment the task it’s built for comes up. You don’t change tools, you don’t pick it from a menu, you just ask for the thing and the right Skill quietly does the work in the background.
Here’s the way I think about it. A prompt is the start of a conversation. A Custom GPT or Gem is a specialist you go and visit. A Skill is a specialist who already knows your business and shows up exactly when you need them, right where you’re already working.
Why Skills matter more than you’d think
This is where it gets genuinely useful for the way you run your business.
The real magic of a Skill is consistency. Most of us don’t have a marketing problem so much as a repeatability problem. You can write a cracking social post when you’re in the zone. The trouble is doing it again next week, and the week after, to the same standard, when you’re busy and tired and just want it off your plate. A Skill holds that standard for you, so the quality doesn’t wobble depending on your mood or your morning or who is using the skill (yes, you can share skills with your team members and take those tasks off your plate #winning).
It also saves you a heap of time. The instructions are already written, so you’re not re-explaining the task or fixing a half-right result. You ask, it delivers, you tweak. Done.
And here’s the part I care about most because it’s central to how I teach AI. A Skill is something you build and own. When you write your own Skill, you understand exactly what it’s doing and why, which means you can direct it, adjust it, and trust it. That’s a world away from using a Custom GPT you stumbled across on the internet and hoping for the best. Skills create freedom. Borrowed shortcuts create dependency.
That fits squarely into what I bang on about constantly: the goal isn’t collecting more AI tools, it’s getting genuinely good at directing the ones you’ve already got. A Skill is one of the clearest ways to turn “I figured out how to do this once” into “my AI does this brilliantly for me, on demand, forever.”
Where you’ll actually bump into Skills
Skills started life in Claude, which is my main tool, so that’s where you’ll meet them most directly. But here’s the exciting bit, and the reason Skills are worth your attention now rather than later, is because they’re quickly becoming a shared standard right across the AI world. That makes them genuinely portable, which is rare and rather wonderful.
Claude and Claude Cowork
This is Skills in their truest form, and where they began. In Claude, you can build a Skill as a SKILL.md file and Claude will use it automatically whenever the task it’s designed for comes up. In Claude Cowork (the version that can work with the files and folders on your computer), Skills become really powerful because they can carry out proper multistep jobs for you, not just write a paragraph. This is exactly the kind of thing I’ll be teaching in my next AI MegaClass.
And now, well beyond Claude
Anthropic (the company behind Claude) didn’t keep Skills to themselves. In late 2025 they published the Skill format as an open standard, meaning any AI platform can adopt it. And adopt it they have, fast. Skills now work in ChatGPT (only for business teams at the time of writing), in Google’s tools, in agent platforms like Manus, and across a growing list of others. Industry-watchers are already calling it the new best-practice standard for this kind of thing, much like what happened a year earlier when Anthropic opened up the technology that lets AI tools connect to your other apps.
What does that actually mean for you? Portability. Build a Skill once, and you’re increasingly able to use that same Skill across different tools, rather than being locked into one. The work you put into capturing how a task should be done isn’t trapped inside a single platform. As more tools adopt the standard (and it’s looking like most of them will), your Skills travel with you. That’s a smart place to be putting your effort.
How this sits alongside Custom GPTs and Gems
You might already have a Custom GPT in ChatGPT or a Gem in Gemini. These are close relatives. Saved, purpose-built assistants you set up once and return to. The key difference is that they tend to live inside one platform, whereas a Skill (built on the open standard) is designed to be far more portable. Tellingly, the platforms are converging on the Skill approach.
ChatGPT has rolled out its own Skills feature. The direction of travel is clear. And if you’ve created a Custom GPT of Gem before, it’s super easy to convert it into a skill. Just take your knowledge and instructions from the GPT or Gem over to Claude and ask Claude to convert it into a skill for you. Magic!
Your wider AI Dream Team
This is really what Skills, Custom GPTs and Gems all add up to. A personalised team of AI helpers, each one trained for a specific job in your business. One that writes your emails in your voice. One that turns a blog into social posts. One that drafts your ad copy. You build them once, and they show up ready to work every time. Build them as Skills, and that team can increasingly come with you wherever you go.
The thread running through all of these is the same. The more clearly you capture how a task should be done, the more reliably your AI does it. A Skill is simply the cleanest, most future-proof way to lock that in.
What sorts of tasks make a great Skill?
Not everything needs to be a Skill. The sweet spot is any task you do regularly, that follows a repeatable pattern, and that you want done to a consistent standard. A few that work beautifully for service-based business owners:
- Writing social posts in your brand voice using your preferred structure
- Turning one blog post into a week’s worth of social content
- Drafting email newsletters that actually sound like you
- Repurposing a podcast or video into written content
- Creating ad copy from a simple brief
- Converting meeting transcripts into accurate minutes with action items
The pattern to look for is simple. If you’ve ever thought “I wish I could just clone the version of me that does this really well,” that task is a strong candidate for a Skill.
What doesn’t make a good Skill? The one-off, the highly creative, or the genuinely strategic. Use Skills for the mundane, repeatable jobs. Keep the masterful, judgement-heavy work in your own hands. That’s the whole point: free yourself from the repetitive stuff so you’ve got more headspace for the work only you can do.
The honest drawbacks of Skills
I’m not going to pretend Skills are a magic wand because they aren’t, and you deserve the full picture.
- They take a bit of effort to set up. You’re investing time upfront to write good instructions. The payoff is real, but it isn’t instant. You’re building an asset, not flicking a switch.
- A Skill is only as good as the thinking behind it. Vague, woolly instructions give you vague, woolly results. The quality of what comes out depends entirely on the clarity of what you put in.
- They still need a human in the loop. A Skill gets you a strong, consistent draft. It doesn’t get you a finished product you should publish unread. You’re still the editor, the brand guardian, and the final say.
- They’re not “set and forget” forever. As your business shifts, or the tools update, your Skills will need the occasional tidy-up to stay sharp. But the good news here is that it’s not that hard to update. Just tell Claude where the Skill needs improvement, and it can make the adjustments for you.
The simplest way to hold all this in your head is to remember that a Skill is a brilliant way to bottle your best work and repeat it. It isn’t a way to remove yourself from the work entirely. You stay in charge. The Skill just does the majority of the legwork, consistently and to your standards.
How to get started with Skills (without the tech headache)
Good news. You don’t need to code, and you can start far more simply than you might think.
- Begin with one task you do all the time. Don’t try to build ten Skills at once. Pick the single repetitive job that drains you most and start there.
- Write down how you do it well. Open a plain document and describe the steps, the rules, the tone, the format. If you can explain it to a new team member, you can write a Skill. That description is basically your first draft. And don’t forget to feed the skill with whatever core knowledge it requires, either by embedding that content as a Markdown file or by linking to the required knowledge in a shared folder in Claude Cowork (if that’s where you intend to use the Skill).
- Let your AI help you build it. This is the lovely bit. You can ask Claude to help you turn your thoughts or notes into a proper Skill. It speaks the format fluently, so you can stay hands-off on the technical side and focus on the thinking.
- Test it, then tweak it. Run the Skill on a real task, see where the output drifts from what you wanted, and sharpen the instructions. A couple of rounds of this, and you’ll have something genuinely reliable.
Start small, start with something annoying and repetitive, and let the tool do the fiddly part. But I will warn you, once you get started making Skills, it is highly addictive. You’ll have an AI Dream Team created in no time at all.
Frequently asked questions about Skills
What is an AI Skill in simple terms?
It’s a saved set of instructions that teaches your AI to do a specific task the same way every time. Rather than re-explaining a job whenever it comes up, you write the instructions once and your AI follows them on demand, giving you consistent results without the repetition.
Are Skills the same as prompts?
Not quite. A prompt is a one-off instruction you type in the moment. A Skill is the full, saved set of instructions for a task that your AI can trigger and follow whenever that task comes up. Think of a prompt as a single conversation and a Skill as a trained team member who already knows the job.
Do I need to know how to code to create a Skill?
Not at all. Skills are usually written in Markdown, which is plain text with a few simple symbols, and you can ask your AI to do the technical formatting for you. Your real job is the thinking: describing clearly how the task should be done. The tool handles the rest.
Which AI tools use Skills?
Skills started in Claude, including Claude Cowork, where they’re at their most powerful. In late 2025 the Skill format was made an open standard, and it’s been adopted quickly, so Skills now work in ChatGPT (including for business teams), Google’s tools, agent platforms like Manus, and a growing list of others. That’s the real beauty of them… Build a Skill once, and it’s increasingly portable across the tools you use, rather than locked into one.
What’s the difference between a Custom GPT or Gem?
They’re close cousins, and the main difference is how you reach them. A Custom GPT or Gem is a separate assistant you switch into and work inside whenever you need that job done. A Skill lives inside the AI tool you’re already using and triggers itself automatically when the relevant task comes up, so you don’t have to leave what you’re doing or pick it from a list. In short, you go to a Custom GPT or Gem, whereas a Skill comes to you.
Skills are how you bottle your best work
Here’s what I really want you to take away. Skills aren’t a developer’s plaything you need to feel intimidated by. They’re one of the most practical ways to get AI doing real work for your business, to your standard, without you having to be there every time.
The shift is subtle but powerful. You stop being the person who does the repetitive task, and you become the person who trains the thing that does it. You stay firmly in charge. AI just becomes a far more capable partner. Human-led, AI-powered, exactly as it should be.
If you want to see this in action, this is precisely what I’m teaching in my next AI Marketing MegaClass on Claude Cowork and Skills, happening on 19 June 2026. It’s a hands-on session where I’ll show you how to build Skills that actually work for your business, so your AI stops guessing and starts delivering.
Grab your spot here: impactiv8.com.au/ai
You’ve got this!
