a trailing comment on a frontmatter key defeats every blank spelling the record_schema gate refuses because the shared same-line scanner strips no comments

a trailing comment on a frontmatter key defeats every blank spelling the record_schema gate refuses because the shared same-line scanner strips no comments

Found by the round-5 ruthless review of the fix for iss-2608301649337965, which closed three of the four spellings that record enumerates and left the fourth.

frontmatter.Fields is a same-line scanner that strips no comments, so the value it hands every gate is the raw remainder of the line. isAbsentValue anchors each of its emptiness tests on the LAST byte — isNull compares the whole string, and isEmptyFlowCollection requires the closing bracket at the end — so a trailing comment defeats all of them at once:

grounds: {}  # todo
grounds: []  # todo
grounds: ~ # todo
grounds: "" # todo

Probed end to end through Lint: each draws ZERO record_schema findings on an admission that is otherwise well formed. admittedProposals never reads grounds, so it keys (rdg-1, rdi-2) as admitted and rdi-2 drops out of report.Undispositioned — the exact mechanism iss-2608301649337965 describes, reached by a fourth spelling.

It is NOT scoped to grounds, and that is what makes it a scanner question rather than a predicate one: severity: minor # todo reads as the value minor # todo to the enum leg, and every same-line scalar in every store is read the same way.

Remedy is a comment strip in the ONE scanner, and it is not a one-liner: YAML starts a comment only at a # preceded by whitespace and outside quotes, so a naive split truncates a legitimate value containing a hash. It belongs where the scanner lives, not in isAbsentValue, or the same escape stays open for every gate that reads a value some other way.

Grounds

• pursued: we expect a comment stripped in the shared scanner to close the escape for every gate at once, rather than for the one predicate the defect was reported against; a value containing a legitimate hash that the scanner truncates, or a commented blank that still reaches a gate as a value, would show it wrong.