GGHL Command  Member Hub Design mockup
22px
DW
The ModulesSet Up Your Command CenterBuild the Command Center (the shared brain)
Video · placeholder

Build the Command Center (the shared brain)

Lesson 3 of 7 · Module 01 · Set Up Your Command Center

This is the lesson where it clicks.

You have the layout from Lesson 2: one project, one folder, all under one roof. Now we build the thing that makes those folders act like one brain instead of a pile of strangers. It's called the Command Center, and you build it by pasting a single prompt and watching Claude do the work.

Here is what gets built, and what each piece is for.

STATE.md is the source of truth for what is happening right now, everywhere. It's one table: workstream, who owns it, status, what it's blocked on, when it was last touched. Every session reads it on the way in and updates it on the way out. This is the file that ends the contradictions. When one session knows a pipeline got renamed, every session knows, because it's written down in one place instead of trapped in a chat you closed.

PROJECTS.md is the registry. One row per project: name, type, purpose, status. It's how Claude knows what exists without you re-explaining your business every morning.

shared/ holds the facts that are true across all projects. ids.md is the single copy of your important IDs and accounts, GHL sub-accounts, Stripe, the rest. The rule is simple: if an ID is in here, trust this copy over anything Claude finds floating in a project. agents.md is where you define any named roles you use.

log/ is the diary. Each session writes a dated entry of what it did, so tomorrow's session can catch up by reading instead of by asking you.

outbox/ is the mailbox. When work belongs to a different session, Claude doesn't reach across and do it. It drops a ready-to-paste prompt in here and you relay it. That's the message bus, and we go deeper on it in Lesson 5.

There's also templates/ and bin/, the reusable pieces and a small script that registers projects. The prompt builds all of it.

Be honest about what this is and isn't. It is not a robot running your agency while you sleep. It is shared memory plus a contract, and right now you are the one relaying prompts between sessions. That's fine. That's the foundation. Everything we build later sits on top of it.

Open the Prompts Pack, copy prompt 01, paste it into Claude Code, and tell it where your projects live when it asks. In about twenty minutes you'll have the whole hub standing.

Run it, then come to Lesson 4, where we teach Claude to read this brain on its own, every session, without you reminding it.

Jerry

Video script

Target: 90-120 seconds spoken. Talking head, on-screen text in [brackets].

On screen: This is where it clicks

This is the lesson where it clicks.

You've got the layout from Lesson 2. One project, one folder, all under one roof. Now we build the thing that makes those folders act like one brain instead of a pile of strangers. And you build it by pasting one prompt and watching Claude do the work.

On screen: command-center/ tree: STATE.md, PROJECTS.md, shared/, log/, outbox/, templates/, bin/

Here's what gets built, and what each piece is for.

On screen: STATE.md: what's happening right now, everywhere

STATE dot M-D is the source of truth for what's happening right now, across every session. One table: the workstream, who owns it, its status, what it's blocked on. Every session reads it going in, updates it going out. This is the file that ends the contradictions from Lesson 1. One session knows a pipeline got renamed, so every session knows, because it's written down instead of trapped in a chat you closed.

On screen: PROJECTS.md: the registry of what exists

PROJECTS dot M-D is the registry. One row per project. It's how Claude knows what exists without you re-explaining your business every morning.

On screen: shared/: facts true across all projects. ids.md = trust this copy

The shared folder holds facts that are true everywhere. Your IDs and accounts live in one file, and the rule is simple: if an ID is in here, trust this copy over anything Claude finds floating around.

On screen: log/ = the diary | outbox/ = the mailbox

Log is the diary. Each session writes what it did, so tomorrow's session catches up by reading instead of asking you. Outbox is the mailbox. When work belongs to a different session, Claude doesn't reach over and do it. It drops a ready-to-paste prompt in here and you relay it.

On screen: Shared memory + a contract. Not a robot running your agency.

Be honest about what this is. It's shared memory plus a contract. It is not a robot running your agency while you sleep. Right now you're the one relaying prompts between sessions, and that's fine. That's the foundation everything later sits on.

On screen: Prompt 01 → paste into Claude Code → ~20 minutes

Open the Pack, copy prompt 01, paste it into Claude Code, and tell it where your projects live when it asks. About twenty minutes later, the whole hub is standing.

On screen: Run prompt 01 → then Lesson 4

Run it, then come to Lesson 4, where we teach Claude to read this brain on its own.