The agent register says intent-auditor serves `intent` alone, while the build loop composes its request and validates its verdict

The agent register says intent-auditor serves intent alone, while the build loop composes its request and validates its verdict

A premise contradiction (severity nitpick) found by the cross-document consistency pass.

• End A: .abcd/development/brief/04-surfaces/README.md:417 — "intent | [05-intent.md] (05-intent.md) | Audits a shipping intent's promise against the delivered code"
• End B: .abcd/development/brief/04-surfaces/34-build.md:697 — "the auditor's fidelity verdict checked against the request (its receipt, both provenance hashes, every criterion and scope condition)"

README's Serves column names the verb that builds an agent's input or validates what it returns. It lists only intent for intent-auditor, while it lists build and drain for the two reviewers. 34-build has the validate stage hand the closing lane to an intent-auditor and check its fidelity verdict against the request before the loop records it. By the column's own definition, build serves the auditor as well.

Evidence: .abcd/work/reviews/2026-10-26-consistency/00-summary.md, finding 690.