Scope the Atlas review to how it works, not whether it is worth it - and restate the rule that Claude must call a script, never an MCP directly

Situation

In the Motherbrain / Atlas sync with Tabatha Wilmot, Karl Lowenbjer and Chris Baek, Peter set the terms of his own code review. He told them outright that he would not wade into whether Atlas is valuable - that is a Tabatha-and-sales question - and that what he wants to answer is how it works, how it is architected, how it stores and persists data, how it does security protections, what should run on a laptop versus centralized, and what the scheduler should be. He asked for the scripts and skills that still live only on Tabatha laptop to be pushed into the repo first so he can read both ends of the system, asked for half-sentence direct answers to Marlon open questions rather than more narrative, and committed to a four-to-five-hour deep dive by Saturday evening with a follow-up the middle of next week. When Tabatha described reps closing deals wrongly through the HubSpot MCP, Peter named the cause - Claude writing code at runtime against a system of record - and restated the standing rule that Claude should never call an MCP directly, it should call a script that structurally cannot do that.

Reasoning

Peter separated two questions that were being handed to him as one. The keep-or-revert recommendation he owes Chris Baek is a technical judgment, and mixing in the value question would both exceed his lane and let a value argument substitute for an architecture answer. He also refused the demo - he said he does not go to class, he reads the book - so he asked for the code and the documentation rather than a walkthrough. The MCP point was not a review finding; it was him using a live incident to restate a forward-looking rule he has applied before, and Karl confirmed Atlas is already architected that way.

Additional Context

Peter had a walk with Tabatha earlier that he says is 95 percent of his current understanding of Atlas. Tabatha is departing CIQ at the end of September, which is what makes the bus-factor questions Peter asked - if you and Karl got run over by a train today, can sales still get stalled-deal insights - the operative ones. The existing todo to review the Atlas code and recommend keep-iterating versus revert to HubSpot predates this session.

Observed Evidence

Peter set the review scope explicitly and unprompted, refused an offered use-case walkthrough twice, asked for the laptop-resident scripts to be pushed to the repo as a precondition, and committed to a specific delivery time - I will dig into this thing by Saturday evening, is my commitment to yall.

Matching Patterns

Confidence Breakdown

Reasoning Depth Analysis