Foundation

AI skills — write a method down once, use it everywhere

You keep re-explaining the same things: how reading notes should be laid out, what a meeting summary must contain, which checks a draft has to pass. Or you keep them in a doc of prompts and paste from it, until the one you need is the one you forgot you wrote. A skill is that explanation saved once, as a page of plain text in your folder — called with / or picked when the job fits, in chat, on a canvas card, or on the text you selected.

Download See one skill in three places →

Built in: 44 canvas methods, plus chat and editor skills
A skill is: One Markdown file you can read
Runs on: macOS (Apple silicon) · Windows
Price: Free app · AI plans from $10/mo, or your own key
AI: Included, or your own API key

The skills settings page with tabs for Chat, Canvas and Editor; the Canvas tab lists two of your own skills, the built-in skills and the example library, with Import and New buttons at the top

Rare

One method, applied where the work is

A prompt saved in a chat app only ever works inside that chat box. A skill is the method itself, so the same file can be switched on for the conversation, the canvas and the editor — and each surface does what it is good at.

In chat, it can take several turns, read your documents and search, and the findings point back at the passages they came from.
On the canvas, applied to a card, the branches come back in the shape the method calls for — one per step, an option comparison as a table, a SWOT as four quadrants.
In the editor, it runs on the paragraph you selected and the result comes back into the document as a tracked change, or is only shown, depending on the skill.
Combinations that cannot work are refused up front: a skill that needs several turns or has to search is greyed out for the canvas and the editor, with the reason.

critical-reading.md

name: critical-reading
description: Use when the user wants a critical reading of a document…

Core claims
Evidence and reasoning
Hidden assumptions
Open questions

In chat

/critical-reading

Hidden assumptionThe sample is all early adopterspaper.pdf · p. 4 ↩

On the canvas

Your card

Core claimsEvidence and reasoningHidden assumptionsOpen questions

In the editor

Rests on one survey; the counter-case is never addressed.Accept · Reject

A skill that needs several turns, or has to search, is greyed out for the canvas and the editor — with the reason — instead of failing halfway.

You do not have to write a single one

Most of the value is already on the shelf. Type / and the first letters of a name, or describe the task and let the matching skill be picked.

The canvas comes with 44 built-in skills in six groups, most of them named methods people have used for decades — SCAMPER, six thinking hats, first principles, a red team, an option matrix.
An example library of ten more — a pre-mortem, Porter's five forces, a stakeholder map — sits one click away, and only counts once you install it.
Chat and the editor have their own built-ins, and any of them can be copied and edited, or hidden if you never use it.

Every built-in thinking model, and what applying one produces →

The slash menu open in chat, listing available skills with their one-line descriptions

Describe the method out loud and get a skill back

The people with the most worth writing down are usually the least interested in learning a file format. Type /new-skill and say what you want done differently.

It asks a couple of questions, drafts the skill, and shows you the whole thing — including the one line that decides when it gets picked.
Nothing is saved until you say so, and overwriting an existing skill takes a second, explicit yes.
What comes back is ordinary text you can keep editing — the draft is a starting point, not a black box.

The chat box with the new-skill command attached, asking you to describe the workflow to turn into a skill: when to use it, the steps and what it should produce

It is a file in your folder, not a setting in someone's account

No manifest, no packaging step, no enable switch. If the file is there it counts; move it out and it stops counting.

At the top: what it is called and when it should be used. Below: the method in ordinary prose — the steps, the checks, the shape the output should take. Keep it with one project, or with you everywhere.
Keep a shelf of them without slowing anything down. Until a skill is used, the AI sees only its name and one line about when it applies — the method is fetched when the job calls for it.
Copy a built-in and it becomes yours to change. Delete your copy and the original comes back.
Sharing one is sending a Markdown file. Importing shows the whole text before anything is saved — no gallery, no account.

Thesis/This project

.mindmax/skills/meeting-notes.md ✓ counts

old-checklist.mdmoved out — no longer counts

Your skills/Every project

house-style.mdverify-claims.md

meeting-notesUse when tidying notes from a meeting

The method · read only when this job comes up

Decisions first, not who said what
Every action gets an owner and a date
No date agreed? Mark it “to be agreed”

house-styleUse when a draft is about to go out under the team’s name

verify-claimsUse when the user asks whether a statement holds up

…the rest of the shelf, one line each

Make a built-in yours

verify-claims · built-inverify-claims · your copy

Delete your copy and the built-in comes back.

Share it

meeting-notes.mdWhole text shownSave

One Markdown file. No gallery, no account.

One skill, from the third repeat to the last fix

Meeting notes are the classic case: the same three preferences, explained after every meeting.

Notice the repeat
The third time you write "decisions first, each with an owner and a date — not a narrative of who said what", that sentence is the skill.
Say it once
Type /new-skill and describe it the way you would to a new colleague. Expect a question back, such as what to do when no date was agreed.
Read the draft properly
This is the step worth your attention. What it drafted is what the AI will follow — and the "when to use" line decides whether it gets picked for a page of rough notes.
Decide where it belongs
This project only, or everywhere. Chat for a pasted transcript, the editor for the notes you typed during the meeting.
Use it
Select the rough notes and run it from the editor, and the tidied version arrives as a tracked change. Or type /meeting-notes in chat with the transcript attached.
Fix the line that let you down
Dates went missing? The instruction did not say what you meant. Open the file, change that one line, run it again.

Coming from a prompt doc or custom instructions

Most people already keep their instructions somewhere. Here is what changes when they become skills — and the rows where what you have now is the better fit.

Topic	A prompt doc or a chat app's custom instructionsWhere most saved instructions live today	In MindMax
Getting it into the job	Find it, copy it, paste it — every time.	Type / and its name, or let it be picked from its one-line description when the task fits.
How many you can keep	One block of custom instructions, or a doc you scroll through.	As many as you like for / ; up to 30 one-line descriptions are offered in chat for automatic matching, so a few sharp ones beat a long shelf.
Where it works	Inside the chat box.	In chat, on a canvas card, and on the text selected in the editor.
Where it lives	In an app's settings or a document somewhere else.	A Markdown file in your folder, synced and backed up however that folder already is.
Always-on preferences	Custom instructions apply to every message without being asked — the right place for "answer in British English" or "keep it short".	A skill applies only when it is called or matched. It is for a kind of job, not for how every answer should sound.
On your phone	A chat app account follows you to the browser and the phone.	A desktop app. Your skills go wherever your folder syncs, but there is no mobile app to use them in.
Skills from other agent tools	Many agent tools now use a SKILL.md file with a name and a description.	The same shape drops in as a chat skill. Only the text is read — scripts or other files bundled beside it are not run.

The same shelf, in every part of the loop

Your notes, and the two hundred PDFs you read to write them, sit in one folder — where a single AI works across both. Tools you would otherwise stitch together each hold one half of that.

All three work on the same folder of files — the one you already have →

Where it gets hard

A skill is instructions, not new abilities
It cannot make the AI do something it otherwise could not. It changes how a job is approached using what already exists — reading your documents, searching, writing files. If you were hoping to bolt on a new capability, this is the wrong mechanism.
A few sharp ones beat a full shelf, and the description does the choosing
Automatic matching reads only the one-line "when to use this", and in chat only 30 of those are offered at once. Write it as the situation it applies to, not a summary of what it contains — and retire the ones you never use, because overlapping descriptions make the right one harder to pick.
An imported skill is instructions you are agreeing to follow
There is no code in a skill and nothing executes, but text can still tell the AI to do things you would not choose. That is why importing shows you the whole file first — read it, the same way you would read a script before running it.

Common questions

What is a skill?

A page of plain text describing how a particular kind of job should be done — the steps, what to check, what the output should look like — with one line at the top saying when it applies. It sits in a folder, and putting it there is what registers it. There is no manifest, no install step and no enable switch.

Do I have to write any?

No. The canvas has 44 built-in skills, most of them named classic methods, and chat and the editor have their own. Type / to see them. When you do want one of your own, type /new-skill and describe the method: it interviews you, shows you the full draft, and saves it only once you approve.

How is this different from custom instructions in a chat app?

Custom instructions apply to every message and live in that app's settings. A skill applies to one kind of job, is called by name or matched from its description, runs in chat, on the canvas and in the editor, and is a file in your own folder. For a standing preference about how every answer should sound, custom-instruction style settings are the better fit.

Can I use SKILL.md files written for other agent tools?

The format is the same shape — a SKILL.md with a name and a description at the top and the method below — so dropping one into the skills folder makes it a chat skill. Only the text carries over: scripts or other files in its folder are not run, and steps that rely on tools MindMax does not have will not happen.

How do I share a skill with someone?

Export it as a single Markdown file and send it however you like. Importing shows you the whole text before anything is saved, because a skill you install is instructions you are about to let the AI follow.

Does a skill let the AI do new things?

No, and that boundary is deliberate. A skill is instructions, not code. It changes how the AI approaches a job using what it can already do — read your documents, search, write files — rather than granting it new abilities.

Write down the thing you keep re-explaining.

Download Pricing

Or start with the 44 that are already there. Skills are plain text in your own folder — uninstall the app and they are still readable.