The personal-insurance delivery-driver proposed mockup is rejected and should not be published. It used a self-generated vertical hero trust strip, did not reliably center headers, used factually overbroad insurance-denial headings, and relied on decorative ordinal cards like Trigger 1 and Step 1 instead of useful visual comparisons.
Installed the missing active Codex skill pack in the Projects skill root:
Updated the existing interactive-element-library skill to block the specific visual failure: decorative ordinal cards, too-small or meaningless eyebrows, and repeated generic trigger or step cards where a comparison ledger, table, lane diagram, or tighter prose is better.
Created a local handoff report: https://www.notion.so
The rejected mockup was not just a spacing problem. It mixed template drift, factual overstatement, and weak design decisions. The skill gap was real: the canonical SA template, web-content revise, brand voice, and pre-publish QA skills existed in repo-specific Claude locations but were not installed in the active Projects skill root for Codex to load by name. The new bridge skill requires the correct load order before future site content work.