the assembler's key and heading signals are scoped to markdown so a record-shaped document inside a Go fixture carries its Audit Notes into the bundle while the manifest asserts refusal

the assembler's key and heading signals are scoped to markdown so a record-shaped document inside a Go fixture carries its Audit Notes into the bundle while the manifest asserts refusal

Found by the itd-183 fidelity audit (rcp-139d9b40f0e8), which ran the instrument rather than reading it.

This is the tenth floor record and the only one that is LIVE ON THE CORPUS. The other nine are recognition gaps on spellings no committed record writes. This one fires today, on this repository, at HEAD.

The key and heading signals are scoped to markdown (project.go:57). So internal/core/site/fixture_test.go — a Go file holding record-shaped fixture text — carries three literal ## Audit Notes sections, WITH THEIR ACCEPTANCE ROLLUPS, into the bundle at item itm-0736, while the same run's manifest asserts that Audit Notes was refused. Manifest and bundle disagree, and the manifest is the side that is wrong.

Mitigating, and why this is major rather than critical: the content is synthetic fixture text, not a real audit; the bound is disclosed at .abcd/development/readings/README.md:73; and no reading runs this cycle, so nothing consumes the bundle. The auditor was explicit that "no reading runs this cycle" bears on the RISK, not on the verdict — the trigger is "when the assembler runs", and it ran.

Why it belongs to the floor intent rather than to a fix here: the class is the same one the other nine share — the floor recognises a construct by the shape of the file it sits in rather than by what the file CONTAINS. A .go file can hold a record; so can a .json fixture, a testdata blob, or a doc comment. Narrowing to markdown is a scope decision that the include table makes and the floor inherits without stating.

Remedy is a design question, not a patch: either the include table admits only what the floor can read, or the floor reads what the include table admits. Seed material for the exclusion-floor intent, and the strongest single argument in it — because it is the one leak a maintainer can reproduce today.

Grounds

• pursued: we expect moving the scope decision onto the table row, and marking every unexamined item in the manifest, to close the container-shape class where widening the parse would have been the other branch, because the manifest then states a fact about each item rather than a claim about the run; an item the manifest marks parsed that no scan ran over, or an unparsed item arriving with no mark, would show it wrong