ba54e13, byte-exact inputChromo Shift PR #8 — candidate status
Evidence preview of commit
ba54e138 · generated 2026-11-15 · view-only. No engine source is served on this page; the interactive engine is not running here (it is domain-locked to production).Build identity (verified)
hs-e2ddc49033Clean rebuild of
12 / 12
npm test pass, 0 failba54e138PR #8 head (draft) — target commit
hs-5770d45c46Earlier report — explained: +1 trailing newline per source file (same code, different bytes)
Footage audit (run1 & run2, 600 frames each)
| Metric | Run 1 | Run 2 |
|---|---|---|
| Frames | 600 | 600 |
| Palettes | 10 | 10 |
| Brand retention (overall) | 99.33% | 99.33% |
| Engine↔pixel agreement | 98.83% | 98.83% |
| Ink-decision switches | 36 | 38 |
| Pass (gate) | FAIL — Mint only | FAIL — Mint only |
9 of 10 palettes retained 60/60 in both runs (Neon, Sunburn, Orchid, Cobalt, Gold Rush, Hot Coral, Ultraviolet, Lime Voltage, Harpstar Rose). The harness is noisy (36 vs 38 switches, earlier 38–40 on the promo); treat counts as samples, not ground truth.
Known issue — Mint Signal (frames 925–928). Both independent runs pick the same washed-out ink
#9AFFCE → #9BFFCF at the same 4 frames, confirmed on the visible pixels (engine↔pixel agree).
Retention there: 56/60 (93.33%). The cause is under investigation (read-only) and the minimal repair is not yet applied or approved.
What this means
Brand-colour and outline behaviour hold on 9 of the 10 promo palettes at the candidate commit; the one reproducible defect is Mint's wash at frames 925–928. No engine change, no merge, no deployment to production has been made (you are looking at a non-`engine` evidence preview).