THE SHORT ANSWER
Yes, for the finite set of configuration keys that move cost per task. Lesson 05 of Becoming Forward Deployed Engineer ships a scope policy (editable paths, forbidden paths, no deletes, no new dependencies, a file cap, forbidden strings) and a guard that caught 6 of 6 mechanical violations and 0 of 3 changes that did more than was asked. One of the three raised top-k from 5 to 20 inside an allowed file, which changes what every question costs. Cost-bearing keys are nameable: the multipliers from the course's own lesson 04 are steps per task, context per call, retries, and caching, plus model choice. List them in the policy, report any added line that sets one of them as a re-quote, and route it to the name on the handshake's cost line. That takes 6 of 9 to 7 of 9. The restructure and the removed validation stay a person's job, and the second belongs in a test, not a diff guard.
The lesson's claim is that the three misses are not a gap you close with another check, because no mechanical rule can compare what was asked with what arrived. For two of them that holds. The third is a cost change, and cost changes are the one class of scope creep whose vocabulary is finite.
I know the shape of the miss from the other side. A four-agent pipeline I approved as "let's split this into specialists" cost 12 times a single agent. The keys that carried that bill had names, and nobody had written them down.
The seventh field, the routing rule, and the honest tally of what stays human are in The Scope Guard Caught 6 of 9. Name the Keys That Carry the Bill. The three-layer rule it rests on is Same Model, Twice the Bill: The FDE Codebase Contract.