content-explorer/workplans/evidence/CONTENT-WP-0005-Review.md

CONTENT-WP-0005 review

Date: 2026-10-19

T01–T08 are done. The public suite is .forgejo/workflows/synthetic.yaml, which runs sh scripts/ci-synthetic.sh and does not fetch OnCo data.

Local CI result: /tmp/grok-goal-60054ed6e434/implementer/ci-local.txt runs the workflow commands (pip install -r requirements.txt, npm ci in ui/, then sh scripts/ci-synthetic.sh). The same suite was run twice before that and both exited 0 (tests-1.txt, tests-2.txt): 26 unittest tests, 34 Vitest tests. Fixture validation and the third-party notice check are in fixtures.txt.

Both example dataspaces pass browse, search, record open, neighbour follow, source and license, and graph focus. The keyboard pass in a11y.txt reached search, a result, the record, a relation link, and return at 1280×800 and 390×844, with page errors 0. Screenshots wide.png and narrow.png show the software-architecture concept index. They are not blank. Their pixels differ, and OnCo pixel equality was not measured (screenshot-compare.txt).

R7 deviation: the keyboard walk, the two screenshots, and the lab navigation timings in vitals.txt (ttfbMs 445.1, fcpMs 828, lcpMs 828) approximate usability. They are not field Core Web Vitals and they are not a usability proof. Screenshot comparison records the wide and narrow layout difference. It does not fail on unequal pixels.

The quality report over the published examples is complete, with every failure class at zero, including missing license, dangling reference, and duplicate identity. The same function reports those three as failures when the input contains them, and publish_directory refuses to write the snapshot. Upstream dispositions are in research/UXParityMatrix.md. Notices still name Copyright (c) 2026 OnCo, the MIT license, and pin b594d7433bf4e408b44215e25e5304696d6f840e.