Module 2 of 6

Prime your Hub

The rhythm already runs (that was M0). Now make it know you. The single biggest thing separating output you can use from output you have to rewrite is how well your Hub understands your role, your goals, and your voice. In this module you teach it all three, and the difference shows up in everything it writes from here on.

By the end of this module

Run /bluerock:onboard and turn what ChatGPT or Claude already knows about you into a working profile, in minutes
Have a CLAUDE.md that knows your role, your work, and your domain, loaded into every session automatically
Own a voice.md style guide so drafts, emails, and recaps sound like you wrote them
Set objectives.md so your morning brief leads with what actually matters this quarter

Watch

The two videos

recording soon

Concept video2–4 min

Why priming compounds: how knowing you makes every downstream skill better, and why voice is the highest-leverage thing to capture.

scripts/M2-prime-concept-script.md

recording soon

Build With Me video5–8 min

Running /bluerock:onboard end to end from Linda’s workspace: the portability prompt, pasting it back, and sharpening the voice guide.

scripts/M2-prime-build-with-me-script.md

Concept

A Hub that knows you compounds

Every skill you will ever run reads your standing context first. So the quality of that context sets a ceiling on everything downstream. Prime it well and the effect compounds:

know you better  →  better output  →  you use it more  →
   more captured context  →  know you better still

That loop is why priming is the highest-leverage hour in this course. The three files you write here are the substrate the rhythm runs on:

File	What it carries
`CLAUDE.md`	Your standing brief: who you are, how to help you, your domain context. Loads every session.
`voice.md`	Your style guide. Every content skill reads it so output sounds like you, not generic AI.
`objectives.md`	Your ranked priorities. daily-brew reads it to decide what counts as focus vs. noise.

The fastest start: your AI already knows you

You have probably used ChatGPT or Claude for months. That assistant has seen how you write and what you work on. /bluerock:onboard starts there: run this prompt in the AI you already use, then paste the result back into /bluerock:onboard.

Based on everything you know about me from our past
conversations and your memory, write a profile I can use
to set up a new AI workspace. Pull from real patterns in
how I've actually worked with you — not generic guesses.
If a section lacks signal, say so rather than inventing.

## Who I am
## What I'm working on
## How I write and communicate (quote 2-3 phrasings)
## How I like AI to help
## Domain context (jargon, tools, people that recur)

No history to mine? /bluerock:onboard still works: it interviews you for the gaps and asks for a couple of writing samples. Either way, you end with the same three files.

Your profile is yours. CLAUDE.md, voice.md, and objectives.md are plain files in your own repo. Nothing is sent anywhere. The most personal context in this whole system lives in files you control and can edit by hand any time.

Worked example

What a good voice guide looks like

Voice is the artifact that earns its keep fastest, so it is worth seeing a good one. Below is examples/voice.example.md from your starter Hub — a sample for a fictional RevOps lead. Read it asking: could an agent write in this person’s voice from this alone? Every example file is browsable in the library.

examples/voice.example.mdworked example

## Tone
Direct and warm. I'd rather be plain than polished.
I say the thing, then stop. No hedging, no cheerleading.

## Sentence length and rhythm
Short. One idea per sentence. I break a long thought
into two rather than run a comma splice.

1Specific enough to act on, not adjectives.

“Direct and warm” alone is a horoscope. The lines that do the work are the rules underneath: say the thing then stop, one idea per sentence. An agent can follow a rule. It cannot follow a vibe.

## Words and moves I avoid
No "leverage", "unlock", "synergy", "circle back".
No exclamation points. No emojis. Never open an email
with "I hope this finds you well."

2The "avoid" list carries the most signal.

Telling an agent what not to do removes the exact tells that make writing sound generated. A sharp avoid-list is often worth more than the tone section, because it kills the defaults the model reaches for.

## Phrasings that sound like me
- "Net: the ICP refresh is working — up 30% on last month."
- "Two things before Thursday's call, then I'll get out
   of your way."
- "Flagging this now so it's not a surprise later."

3Real quotes are the ground truth.

Two or three things you have actually written teach voice better than any description. The model pattern-matches against them. This is exactly what /bluerock:onboard pulls from your writing samples — and what you sharpen by hand afterward.

The takeaway: a voice guide is rules plus real examples, not adjectives. That is what makes output sound like you.

You build

Prime your own Hub

Your turn

Teach your Hub who you are. Twenty minutes of this pays back in every draft, brief, and recap you generate from here on.

1Pull your context. Open the AI you already use and run the portability prompt from the Concept section above. Copy its answer. (No history? Skip to step 2 and let /bluerock:onboard interview you.)
2Run /bluerock:onboard. Paste the profile in, and add two writing samples you are proud of — a post, an email. It will ask a few questions where it is thin, then draft CLAUDE.md, voice.md, and objectives.md.
3Sharpen the voice guide. Open voice.md and compare it to examples/voice.example.md. Tighten the avoid list and the “phrasings that sound like me” quotes — those two sections carry the most signal.
4Rank your objectives. Open objectives.md and make the top three specific enough that an agent could name your #1 back to you. Park what you are not doing, so the brief stops surfacing it.
5Test the voice. In a new chat: Read voice.md, then draft a two-line note declining a meeting. Does it sound like you, or like a brand? If it is off, the fix is in voice.md, not the prompt — edit it and try again.
6Make it durable. Source Control panel → message → ✓ Commit → Sync Changes. Your profile is now saved to GitHub. From here, every skill you run reads it.

You are done when

Checkmarks save in this browser only

The tell that it worked: ask daily-brew for a brief tomorrow and watch it lead with something from objectives.md instead of generic “keep making progress.” Priming is what turns the rhythm from running to useful.

Use it for real

Between now and Module 3

Take-home

Use your primed Hub for one real piece of writing this week — an email, a Slack update, a short brief. Generate it, then notice where it sounds like you and where it does not, and fix voice.md accordingly. A voice guide gets sharp through use, not in one sitting.

Before the next module

Checkmarks save in this browser only

The meta-layer

How Linda does this

Patterns from How I work with AI that show up in M2:

3Markdown is memory

Your whole profile is plain markdown: legible to you, parseable by the agent, versioned by the repo, owned by you.

4Voice is encoded, not hoped for

A style guide is the difference between output that sounds like you and output you have to rewrite. You write the rules down once.

5Describe, don't hand-edit

You don’t fill the brackets by hand. You describe yourself and let the agent file it — the whole curriculum in miniature.

M1 — Anatomy of an agent

M3 — Skills as operational playbooks