Grouping issue, filed as part of a catalog pass over the 2026-08-28 to 2026-08-31 resync and MTP conversion campaign. This issue records the shared cause and adopts the affected issues as sub-issues. It proposes no fix.
The shared cause
verbatim fidelity means a downstream copy is byte-identical to the hub canonical, and spec/audit.py reports any difference as drift. The tier has exactly one setting, so it assumes that every byte of a carried file or section is equally correct in every carrier. Two different findings show that assumption failing, in opposite directions.
A carrier needs its own value, and the tier has no way to give it one. #1118 finds .markdownlint-cli2.jsonc declared verbatim while being the single source of truth for what markdownlint lints in that repository. A repository whose tree holds committed data rather than authored prose needs its own globs and ignores. There is no per-repo mechanism, so it either fails the verbatim check permanently or loses real lint scoping on the next re-vendor.
The hub's own value is wrong in every carrier, and no carrier may fix it. #1129 finds verbatim text carrying hub-local references. A bare #1073 in GOVERNANCE.md renders relative to whichever repository displays it, so it is a dead link in ptr727/PhotoCleaner today and becomes an actively wrong link to an unrelated issue once that repository reaches 1073. A host-setup/ path in repo-worktree/SKILL.md has no hub qualifier. The section byte-matches the canonical, so the audit is satisfied while every carrier is wrong.
The two together bound the problem: the tier cannot express a value that legitimately varies, and it cannot detect a value that is uniformly wrong.
Children
| Issue |
Direction |
| #1118 |
A carrier legitimately needs to differ, and verbatim forbids it |
| #1129 |
The hub value is wrong everywhere, and verbatim forbids fixing it locally |
What a fix at this level would have to address
Whether the model needs a third position between verbatim and intent, something like a carried file with declared per-repo fields, and separately whether verbatim content should be linted for references that only resolve in the hub.
Related but separately caused: #1075 and #1120 are about the declaration drifting from the document, not about what the tiers can express. #1110 is the same question at the intent end and is grouped with the audit's own accuracy.
Grouping issue, filed as part of a catalog pass over the 2026-08-28 to 2026-08-31 resync and MTP conversion campaign. This issue records the shared cause and adopts the affected issues as sub-issues. It proposes no fix.
The shared cause
verbatimfidelity means a downstream copy is byte-identical to the hub canonical, andspec/audit.pyreports any difference as drift. The tier has exactly one setting, so it assumes that every byte of a carried file or section is equally correct in every carrier. Two different findings show that assumption failing, in opposite directions.A carrier needs its own value, and the tier has no way to give it one. #1118 finds
.markdownlint-cli2.jsoncdeclaredverbatimwhile being the single source of truth for what markdownlint lints in that repository. A repository whose tree holds committed data rather than authored prose needs its ownglobsandignores. There is no per-repo mechanism, so it either fails the verbatim check permanently or loses real lint scoping on the next re-vendor.The hub's own value is wrong in every carrier, and no carrier may fix it. #1129 finds verbatim text carrying hub-local references. A bare
#1073inGOVERNANCE.mdrenders relative to whichever repository displays it, so it is a dead link in ptr727/PhotoCleaner today and becomes an actively wrong link to an unrelated issue once that repository reaches 1073. Ahost-setup/path inrepo-worktree/SKILL.mdhas no hub qualifier. The section byte-matches the canonical, so the audit is satisfied while every carrier is wrong.The two together bound the problem: the tier cannot express a value that legitimately varies, and it cannot detect a value that is uniformly wrong.
Children
verbatimforbids itverbatimforbids fixing it locallyWhat a fix at this level would have to address
Whether the model needs a third position between
verbatimandintent, something like a carried file with declared per-repo fields, and separately whether verbatim content should be linted for references that only resolve in the hub.Related but separately caused: #1075 and #1120 are about the declaration drifting from the document, not about what the tiers can express. #1110 is the same question at the
intentend and is grouped with the audit's own accuracy.