Copilot's full-diff review of HomeAssistant-Config PR #33 found two carried-document portability gaps against ProjectTemplate 74ef727.
GOVERNANCE.md "Workflow YAML Conventions" links to ProjectTemplate's canonical publish-release.yml. The carried-document rule bans template-repository references except when the reader is told to run a hub-hosted tool. This link is an implementation example, not a tool invocation, so the canonical verbatim section contradicts its own carried-reference rule. The new relative-link validator accepts the absolute URL and therefore does not detect this class.- Carried
AUDIT.md opens by saying the fleet ground truth is "in this repo" and links to registry/, reports/, STANDUP.md, and RESYNC.md. Those paths are hub-only and absent from downstream repositories. Because AUDIT.md is whole-file intent rather than section-verbatim, the new carried-link validator does not inspect it.
Please make both carried surfaces truthful and portable downstream, then extend the new validation so equivalent regressions fail at the hub before re-vendoring.
Copilot's full-diff review of HomeAssistant-Config PR #33 found two carried-document portability gaps against ProjectTemplate
74ef727.GOVERNANCE.md"Workflow YAML Conventions" links to ProjectTemplate's canonicalpublish-release.yml. The carried-document rule bans template-repository references except when the reader is told to run a hub-hosted tool. This link is an implementation example, not a tool invocation, so the canonical verbatim section contradicts its own carried-reference rule. The new relative-link validator accepts the absolute URL and therefore does not detect this class.AUDIT.mdopens by saying the fleet ground truth is "in this repo" and links toregistry/,reports/,STANDUP.md, andRESYNC.md. Those paths are hub-only and absent from downstream repositories. BecauseAUDIT.mdis whole-file intent rather than section-verbatim, the new carried-link validator does not inspect it.Please make both carried surfaces truthful and portable downstream, then extend the new validation so equivalent regressions fail at the hub before re-vendoring.