Workflow 002: The Me Skill

You're missing one thing that would make AI easier every single time you use it.

It doesn't know who you are.

A Me skill is a saved context file describing who you are, what you run, and how you work. Claude and ChatGPT both load it automatically before every task, so the AI starts each conversation already knowing your background, your goals, your voice, and your rules. You build it once, in about twenty minutes, by letting the AI interview you. After that you stop re-explaining yourself and the output stops sounding like it was written for a stranger.

This is the foundation every other workflow in this series sits on.

What this replaces

Re-teaching the AI who you are, in every new conversation, forever.

Before this, I'd get output that was technically correct and completely wrong for me. Too formal. Too long. Written for an audience I don't have. Recommending things that don't fit how I actually work.

Every chat started with three paragraphs of throat-clearing before I could ask for the thing I wanted.

Now the context loads by itself. I ask for a draft and it already knows my voice, my business, my team, and what I'll hate.

Blueprint

Claude and ChatGPT both added skills earlier this year. A skill is a saved instruction file that runs the same way every time it's activated. Think of it as a recipe the AI follows.

The Me skill is one file containing everything the AI needs to know about you as a person and an operator.

Trigger: automatic, before every task. Input: your background, goals, working preferences, voice, and rules. Output: AI that responds like someone who has worked with you for a year.

Here's the part that makes this easy. Both platforms have a skill for creating skills. You are not writing a document from scratch. You are getting interviewed.

Organize

What the file needs to contain:

• Who you are: background, business, role, what you actually do all day
• What you're building toward: your current 90-day goal and the longer one
• How you work: your pace, your energy, what you delegate, what you refuse to
• Your voice: how you write, words you never use, the tone that sounds like you
• Your people: team members, roles, who handles what
• Your non-negotiables: the rules that never bend

If you have ever onboarded an assistant properly, you already have this material sitting in a document somewhere.

Systematize

Do not write this file cold. Get interviewed for it. Open Claude or ChatGPT and paste this:

We're going to create a Me skill: a context file about who I am that you'll
load before every task, so I never have to re-explain myself.

Interview me first. Ask me questions, a few at a time, covering:
- My background and what I actually do all day
- My business, my role, my team
- What I'm working toward right now
- How I like to work: pace, format, level of detail
- My voice: how I write, words I never use, tone that sounds like me
- What I want you to always do, and never do

Ask follow-ups when my answers are vague. Push me for specifics and real
examples instead of accepting general statements.

When you have enough, write the file in the third person, organized in
clear sections. Then show it to me so I can correct it before we save it.

How do you set it up?

1. Open Claude or ChatGPT and paste the prompt above.
2. Answer the questions. Out loud is fine, use voice input if typing is slow.
3. Read the draft it produces and correct anything that sounds like a stranger wrote it.
4. Save it as a skill so it loads automatically.
5. Update it when something real changes. Mine gets refreshed about once a quarter.

What makes the file actually work

Give examples, not adjectives. Not "I like concise writing." Paste something you wrote and say "like this."

Include what you hate. The nevers do more work than the alwayses.

Assign

What the AI does alone: run the interview, organize your answers, draft the file, and apply that context to every task afterward.

What stays human: the answers, and the correction pass. Read the draft before you save it. AI will make you sound slightly more polished than you are, and if that version gets saved, everything it writes for you afterward carries a voice that is almost yours but not quite.

Something I noticed while building mine. This is the exact process we use to onboard an executive assistant to a founder. Background, working preferences, communication style, non-negotiables, who's who on the team. Ten years of running that onboarding for humans, and the questions turn out to be identical for AI.

Which points at the real problem. The bottleneck was never the assistant. It's that most founders have never written down how they work, so every new helper, human or AI, starts from zero.

Rule of thumb from this build: if you can't hand it to a new assistant on their first day, AI can't use it either. Systems before AI. Systems before your next hire.

Integrate

The Me skill is the base layer. Everything else stands on it:

• The morning brief agent workflow knows what to prioritize because it knows the 90-day goal that lives here
• Content workflows sound like me because the voice rules live here
• Every new agent I build inherits this context instead of needing its own version

Build this one first. Everything downstream gets better and you only do it once.

Tomorrow: the Business skill. The Me skill teaches AI who you are. The Business skill teaches it what you run.