five nits from the itd-179 delta review including two exported functions with no caller and a renamed citation that now names the wrong predicate

five nits from the itd-179 delta review including two exported functions with no caller and a renamed citation that now names the wrong predicate

Five nits from the itd-179 delta review, held on one record. Per the round rule adopted 2026-08-30 these are settled at the ship commit, not by re-opening a review.

1. intent/claims.go -- the extraction renamed a correct citation into an incorrect one. It reads mdrecord.IsAnyBullet already rules for acceptance criteria; pre-extraction it cited the TOP-LEVEL predicate, and acceptance criteria are counted by IsTopLevelBullet. IsAnyBullet rules nothing about them.
2. mdrecord.go -- IsAnyBullet is exported with no caller in the tree (its only occurrence is the mis-citation above), and Masked is exported with no caller outside its own package.
3. capture/workflow.go -- appendGrounds is passed the STATE rather than the verb, so one command emits two prefixes: resolve: grounds refused from the flag check and resolved: grounds refused from the append.
4. grounds/record.go -- a record with two ## Grounds headings reads only the first, so the second section's entries are invisible everywhere. The intent half guards the analogous case for scope conditions by refusing to stamp when it counts more than one heading; grounds has no such guard on either side. Pre-existing pattern on new surface.
5. capture/validate.go -- groundsEntries's comment says it asks the same reader the intent half asks. The intent half asks ParseSectionAboveFloor; this asks ParseSection. The next paragraph corrects it, so the block contradicts itself rather than being simply wrong.

Two more from the delta SECURITY review, folded here rather than given ids:

1. internal/core/mdrecord/ ships with no test files of its own and is covered only transitively through core/intent and core/grounds. A new leaf in the core DAG with no direct tests is the shape adr-57 makes load-bearing.
2. opensCommentFrom/findBacktickRun is superlinear on a line carrying many distinct-length backtick runs, because each unmatched run rescans to end of line. PRE-EXISTING, moved verbatim from claims.go; record files are small, so there is no practical consequence today.

Grounds

• pursued: we expect an unused export and a mis-cited predicate to mislead the next reader in the same way a false message does, so both are settled at the ship commit rather than carried