Wire it into your ritual

Lesson 6 of 9 · Module 02 · The Agent Team: Catch What GHL Hides

A check you run when you remember is a check you will skip. The point of this lesson is to make Vera something you do not have to remember, because it lives inside the loop you already built in Module 01.

Back in Module 01 you set up the Command Center: a place for Claude to think, one set of rules, a shared memory to report into, and the read-first / write-last loop. Read the state before you touch anything, write changes last, log what you did. The agents only work because that structure is already there. Now you plug the checks into it.

Two checks, two moments

There are two moments where a workflow can rot, so there are two checks.

The first is audit-before-deploy. This is Vera, and it belongs at the write step. The loop already says write last. The new rule is: the last thing before a write to any workflow is a Vera audit. You read the state, you make your change, you run Vera, then and only then do you deploy. If she blocks, you do not write. The audit becomes the gate the write has to pass through.

The second is the daily sweep. Workflows do not only break when you touch them. A pipeline gets deleted in another session, a snapshot gets applied, a stage gets renamed by a teammate, and a workflow you have not opened in a month is now dead from step three down. The fix is a scheduled sweep that runs across the account on its own and lands a report before you start work. That sweep is the next agent in the build, and Module 02's later lessons cover it once it is proven live. For now, hold the slot for it in your ritual. The daily report goes at the top of your morning read, before you open a single account.

Where the result lands

You run the audit yourself, before you deploy, and the result comes back in your Claude conversation: clean, or blocked with the exact dead reference named. Then you do what Module 01 taught you to do with anything that matters. You log it. The change and the fact that it cleared an audit both go into your Command Center, so the record shows not just what you changed but that it passed a check before it went live.

The daily sweep is the other half, and it works differently: it runs on its own and lands a report before you start work, without you invoking it. That is the scheduled agent still in the build, the one held back in this module until it is proven live. When it ships, the morning report arrives on its own. For now, the audit-before-deploy check is the one that is yours to run, and it is the one that catches the bug at the exact moment it would have shipped.

The principle is the one from Module 01: nothing happens silently. GHL hides the failure. Your Command Center does the opposite. Every check leaves a trail in the place you already look.

The loop, updated

Here is the read-first / write-last loop with the checks wired in:

1. READ    open the daily sweep report first. Start from known state.
2. WORK    make your changes in the right sub-account.
3. AUDIT   run Vera on every workflow you touched. (write gate)
4. WRITE   deploy only what Vera cleared.
5. LOG     record the change, and that it cleared the audit, in your Command Center.

That is the whole habit. You are not adding a tool you have to babysit. You are adding two checkpoints to a loop you already run, and routing them into a channel you already read.

In the next lesson, a look at the rest of the team, who is coming and what they will own.