TOOLS & SKILLS
What's An .md File? (Crash Course) 🛠️
Markdown files sound technical. They're not. They're just plain text AI can read easily, and they're the foundation of every AI agent. Here's the simple.
AI sounds complicated, but I'm going to break down one of the most important pieces so you can actually start to learn it, and don't get left behind. Let me show you what an md file really is, and how to build your first one.
An md file (or markdown file) sounds technical, but it's literally just a plain text file with simple formatting so AI can read it easily. That's all "marked down" means. When you work with AI, it reads files like this to know how to work with you.
Here are the only marks you'll actually use, this is the whole "language":
# Title makes a big heading. More hashes (##, ###) make smaller sub-headings.
- item makes a bullet point. Stack them for a list.
**bold** makes text bold for emphasis.
1. step makes a numbered list, for steps in order.
That's basically it. Plain words plus those four marks, and you can write a markdown file.
Files Are How Agents Think
Here's the part that unlocks everything. An AI agent isn't something super complex either. It's genuinely just a smart assistant that knows how to read the right file at the right time.
So when you have an agent, you'll have a bunch of md files that spell out your preferences, all the information about your business or your job, and any skills or repeatable workflows you have. The agent reads through them like a table of contents, so it knows exactly which one to open when. That's it. That's the whole magic trick.
In plain English
A markdown file is just a note you write for your AI instead of for a person. An agent is just an assistant with a folder of those notes and the sense to grab the right one. Once that clicks, building your own is no longer scary.
Build Your First File
The single most useful file to start with is a "How I Work" file. It is a one-page note that tells any AI who you are, how you want it to talk to you, and what you need help with, so you stop re-explaining yourself in every single chat. This is the exact foundation every AI agent is built on.
There are three ways to make yours: fill in the template below, let AI interview you to build it (the prompt for that is right under the template), or open Claude, turn on voice-to-text, and just ramble, then say "turn this into a clean markdown file." Whichever you pick, the finished file is yours to keep and reuse forever.
Your "How I Work" file (fill in the brackets)
Prefer not to fill in blanks? This is the fastest way to a great file. Paste this into Claude or ChatGPT and it will interview you, then write the whole thing for you in about 10 minutes.
Or let AI build it for you
One Rule Before You Paste
Keep this file to context, not secrets. It is safe to include your role, goals, voice, and business facts. Do not put passwords, card or bank numbers, IDs, customer personal info, or your home address in it. If you would not paste it onto a public website, leave it out.
Now What Do You Do With It?
A file sitting in a note does nothing. Here is how to actually plug it into your AI so it works the way you do. Start with the easy way, then set up whichever tool you use so you never have to paste it again.
Start a new chat and paste your whole file first, with this line at the top: "This is my How I Work file. Use it for everything in this chat. Read it, confirm you have it, then wait for my question." Then ask away. This works in Claude, ChatGPT, Gemini, any of them, with zero setup. Do this one first.
For every chat: go to Settings, then "Instructions for Claude," and paste your file there. Now it applies to everything you do.
Even better, make a Project: go to Projects, click New Project, name it "About Me," click "Set project instructions" and paste your file, then upload the actual .md file into the project knowledge panel on the right. Every chat inside that Project now knows you, and the file stays put.
Go to Settings, then Personalization, then Custom Instructions. You will see two boxes. Put the who-you-are and goals part in the first box, and the how-to-talk-to-me and never-do rules in the second. Each box has a size limit, so paste a shortened version here and keep the full file for pasting or for a Project.
Keep It Alive
Your file gets better the more you use it. Every time you catch yourself re-explaining something to the AI, add that line to the file and update the date at the top. Once a month, skim it and refresh your goals. To check it is working, ask the AI: "Based on my file, how should you talk to me and what should you never do?" If it repeats your rules back, you are wired in.
The Real Win
Once the markdown concept clicks, you understand how to build your own agents, and a whole new world of AI opens up. Your "How I Work" file is step one. Build it once, plug it in, and watch how much less you repeat yourself, and how much more the AI sounds like it actually knows you.
The real unlock of AI agents is not full autonomy, but human-in-the-loop systems design. Structure your agent inputs with explicit constraints, negative prompts, and automated test checkpoints. When building tools, keep token consumption lean by caching system prompts and isolating tasks into specialized sub-agents.