Find where reasoning and authority are collapsed

The prompt

In this repository, list every place Claude Code (or any agent) can currently execute an action directly, without a separate identity check, policy check, or CI gate in between. For each one, name the specific service, IAM principal, log source, and deployment gate that should sit between the model's decision and the action.

Why it's built that way

The handbook's architecture rule: a model may recommend a tool call, but the tool contract, identity, policy engine, network path, and target system determine whether it executes. Mapping the control to a named service and log source is what makes the architecture reviewable and testable — rather than dependent on prompt wording.

What to do with the answer

Turn each gap you find into a tracked follow-up: which existing service (an IAM role, a CI gate, a policy engine) should own that boundary, and who owns closing it.