Skip to content

Markdown Basics

Markdown is the format your AI writes in, and the format nearly everything in this playbook is stored in — skills, prompts, context files, and your AI Registry.

Here is the whole ask: you mostly read Markdown. Your AI mostly writes it. You need to recognize six symbols and know how to save a file. That is the entire skill.

Nothing on this page requires you to install anything.

Every time Claude or ChatGPT answers you with headings, bold text, and bullet points, you are looking at rendered Markdown. The model wrote a handful of symbols; the app turned them into formatting before you saw them.

And every time you have pasted an AI answer into an email and watched stray **asterisks** or ### hash marks come along for the ride — that was the same Markdown, arriving somewhere that did not know how to render it.

So this is not a new thing to learn. It is a thing you have been reading for months, finally named.

Markdown is plain text plus a small number of punctuation marks that mean formatting.

You already do this instinctively. In Slack or email you type *important*, or start a line with a dash to make a list, and you expect it to look like something. (One small trap: in Slack a single asterisk means bold, but in Markdown it means italic — double asterisks are bold.) Markdown formalizes that instinct into a small, consistent set of rules that most modern tools understand.

Three terms that get tangled together:

  • Markdown — plain text with light formatting marks. Readable as-is, even before anything renders it.
  • HTML — the language web pages are built from. Far more powerful, far heavier to write.
  • Rich text — what Word and Google Docs produce. Looks finished, but the formatting is locked inside that application’s own file format.

A Markdown file is simply a text file whose name ends in .md.

1. Your AI reads and writes it natively. Models produce Markdown by default. Handing them Markdown and taking Markdown back is a clean round trip — no reformatting, no copy-paste mangling, nothing lost in translation.

2. Structure is the point. Headings and lists tell the model what is a section, what is a list of rules, and what is an example. A wall of prose does not. This is the direct link to Context: context is what you give the AI; Markdown is how you write it down so the model can navigate it.

3. It is plain text, so it lasts. No corrupted files, no application lock-in, no document that only opens in the program that made it. A .md file written today opens in every tool you will ever use, including ones that do not exist yet.

4. It is what this playbook runs on. Your skills, prompts, context files, and AI Registry are all .md. You will meet Markdown whether or not you go looking for it.

Here is a complete file — meeting notes from a client call. This is what you would actually type:

# Client Sync — Acme Corp
**Date:** 12 March **Attendees:** James Gray, Sara Reyes
## Decisions
- Ship the pilot to three teams first, not the whole org
- **Budget approved** at $40k for Q3
- Sara owns the rollout comms
## Open Questions
- [ ] Who owns the onboarding docs?
- [ ] Do we need legal review before the pilot?
- [x] Confirm the Q3 start date

And here is the same file, rendered:


Date: 12 March Attendees: James Gray, Sara Reyes

  • Ship the pilot to three teams first, not the whole org
  • Budget approved at $40k for Q3
  • Sara owns the rollout comms
  • Who owns the onboarding docs?
  • Do we need legal review before the pilot?
  • Confirm the Q3 start date

Same information, twice. The version on top is what lives in the file; the version underneath is what you and your AI both see. Nothing was lost — and the file is still something you could open in Notepad.

Why the headings look smaller here than in your own file: this page shows them at a reduced size so they do not clutter the contents menu on the right. In a file of your own, # gives you a large document title and ## a clear section heading — both bigger and heavier than what you see above.

Six symbols. Some have two forms — two heading sizes, two kinds of list, ticked and unticked boxes — but there are still only six things to learn.

SymbolYou typeYou get
1. Headings# Big HeadingThe title of the document
## Section HeadingA section within it
2. Bold**urgent**urgent
3. Lists- Ship the deckA bullet point
1. First stepA numbered list
4. Checklists- [ ] Send the recap☐ An empty checkbox
- [x] Send the recap☑ A ticked checkbox
5. Links[our site](https://handsonai.info)our site
6. Tables| Name | Status |A table row (see the note below)

On tables: they are the one symbol that is genuinely tedious to type by hand — every row needs its pipes lined up. Do not bother. Ask your AI: “put that in a Markdown table.” It will get the pipes right every time.

You will also see these: `backticks` around file names and commands, > at the start of a line for a quoted passage, and ![alt](image.png) for images. You rarely need to type them, but now they will not surprise you.

Anything beyond these six is rare enough that the better move is to ask your AI: “how do I do X in Markdown?” — you will get the answer faster than looking it up, and in the context of what you are already writing.

Two routes, and you will mostly use the first.

Ask your AI for it. Add four words to any request: “…and write it in Markdown.” That is the whole technique. The model already writes Markdown by default; you are just telling it to keep the symbols rather than strip them.

Convert what you already have. Paste an existing document into your AI and ask: “Convert this to Markdown, keeping the headings and lists.” Old meeting notes, a style guide, a process document — anything you want to start reusing with AI is a good candidate. If the document is already in Google Docs, you can skip the copy-and-paste entirely — see Turn on Markdown in Google Docs below.

Start here — nothing to install. A .md file is a plain text file, so tools you already have will open it.

  1. Open TextEdit (press Cmd + Space, type TextEdit, press Enter).
  2. Go to Format in the menu bar and choose Make Plain Text. Do this before you type anything — see the gotchas below for why.
  3. Type or paste your Markdown.
  4. Press Cmd + S. In the Save As box, type the full name including the ending: my-context.md.
  5. If a dialog asks about the extension, click Use .md.

Where it renders, and where it does not. This trips people up on their first file, so it is worth knowing in advance:

Where you put itWhat you see
Your AI chat (attached or pasted)Fully rendered — this is the one that matters most
GitHub (open the .md file)Fully rendered
Notion (paste the text in)Converts to Notion blocks as it arrives
Slack (paste the text in)Bold and lists render; checkboxes do not — they appear as literal [ ] characters
Google Drive (open the .md file)Raw text, symbols and all. Nothing is broken; Drive just does not render Markdown
TextEdit / NotepadRaw text — though recent Windows 11 Notepad formats Markdown as you type. Use View to switch back to the raw symbols.

When you are managing more than a few files, move to a proper editor with a live preview pane. That is a setup step of its own — see the AI Code Editor guide.

The question everyone asks: if I write in Markdown, how do I send it to my CEO?

Markdown is where you draft and store. It converts at the last step, and you have three ways to do it:

  • Paste it. Copy the AI’s answer as it appears on screen — the formatted version, not the raw symbols — then paste into Google Docs, Notion, or Slack. The formatting comes with it. (Notion also converts raw Markdown into its own blocks as you paste.)
  • Ask your AI. “Turn this into a Word document” or “rewrite this as an email I can send to my CFO.” This is the most common route by far, and the model handles the tone shift as well as the format.
  • Turn it into a PDF. Paste it into Google Docs (see above), then use File → Download → PDF Document. Or simply ask your AI to produce the file for you.

Nobody has to know your source file was Markdown. That is rather the point.

Open a skill file or a registry entry and you will find something like this at the very top, fenced by three dashes:

---
name: weekly-report
description: Drafts my Monday status update
---

That is called frontmatter. It is a small label on the file — a title, a description, a date — that tools read to know what the file is and how to use it. Your AI uses it to decide when a skill applies.

You almost never write frontmatter by hand; the framework skills generate it. You just need to recognize it and leave it alone. If you delete those lines, tools stop recognizing the file.

This exercise leaves you with a file you will genuinely reuse: a context file describing your role and your company, which you can attach to any future chat so you stop re-introducing yourself.

  1. Ask your AI. Paste this exactly:

    Interview me with three short questions about my role and my company. Then write my answers up as a context file in Markdown, using headings and bullet points.

  2. Answer the three questions.

  3. Copy the Markdown it produces. Copy the raw symbols, not a screenshot — if your AI shows a Copy button on the block, use that.

  4. Save it as a file. Follow the macOS or Windows steps in Viewing and Editing It above, and name it my-context.md.

  5. Edit one line. Open the file, change a single detail — add a project you are working on — and save again. You have now edited Markdown.

  6. Hand it back. Start a brand-new chat. Attach my-context.md using the paperclip or + button beside the message box, then ask for something ordinary: “Draft a short intro paragraph about me for a conference bio.”

Watch what happens on that last step. The AI does not ask who you are. That is the whole argument for Markdown, demonstrated in one file.

Three things go wrong on the first attempt, and none of them are about the symbols themselves.

TextEdit saves rich text by default. If you skip Format → Make Plain Text, TextEdit writes a formatted document instead of plain text, and your symbols will not behave. Set plain text first, before typing.

Smart quotes. TextEdit may convert straight quotes into curly ones. Harmless in ordinary prose, but it breaks frontmatter. Turn it off under Edit → Substitutions → Smart Quotes.

Blank lines matter — on both platforms. Markdown needs an empty line between a paragraph and a list, or between a heading and the text beneath it. Without that gap, things run together. When something looks wrong, a missing blank line is the first thing to check.

Now that you can read it, you will see it everywhere:

  • Context — the background material you give your AI. Markdown is how you write it down so the model can navigate it.
  • Skills — every skill is a Markdown file with frontmatter at the top.
  • Prompts — structured prompts use headings and lists for the same reason your context files do.
  • AI Registry — your inventory of what you have built, kept as Markdown in your workspace.

You do not need to write any of these by hand. But you can now read them, edit them, and tell when something looks wrong — which is the whole job.