Images inside the development journal · mobile fix

Fixed the journal renderer, which previously printed capture references as prose without creating images. Added strict approved-capture references and captions, with responsive cards and the existing provenance dialog. Historical text is unchanged; explicit images were attached only where earlier entries already named the exact capture or session/iteration. Entries without identified visual evidence remain text. Session journal-images-20261014 preserves iteration 00-before (plain paths), 01-image-cards (initial rendering), and 02-curated-captions (final phone light/dark). All four site captures were individually reviewed. Measurement snapshot on 14 October 2026, source source-a431724da629, UI journal-ui-9830e9e9fad1-8d0434bd2644: historical journal images increased from 0 to 16 across 9 entries; all 16 image URLs returned HTTP 200 through private private preview HTTPS; 8 rendered layouts (320, 390, 768 and 1440 CSS pixels in both themes) had zero page/card horizontal overflow; 16/16 renderer regression checks and 65/65 Python overview/export/evidence checks passed in the final focused run at 21:17 UTC, including the real-manifest export smoke. The exported archive preserved all 22 image references across 11 journal entries then present, with byte-identical image files and both images/full measurements retained in the affected entry. Other sessions added entries during this work, so these final archive totals differ from the earlier layout snapshot. Source: rendered browser DOM, HTTPS responses, node tests/test_journal_renderer.mjs, and focused pytest output. The affected mobile contrast entry displays two images; tapping its first card loaded the full image, and Escape closed the dialog. These measurements precede this new entry, whose four evidence cards are additional. A separate broader repository run reported 208 passed and one failure in test_object_use_changes_only_validated_nearby_state[bed-Dreaming] (resident version 3 versus expected 2); the source fingerprint changed during that broader run. This simulator behavior is outside the journal fix and remains unresolved; that run does not establish a stable full-suite result. No physical-phone test was performed by this agent. See overview-site documentation for journal authoring.