What problem does it solve? Golden tests fail whenever an intentional visual change lands, and the reflex response — running flutter test --update-goldens — silently destroys the only record of what the UI used to look like. This Skill enforces the one sanctioned ritual for overwriting committed reference images so a re-baseline never hides a regression, a flaky test, or host-specific rendering noise. ## Core Features & Use Cases - Ordered blessing workflow: Green the non-golden geometry, contrast, and a11y assertions first, then re-baseline only in the pinned OS/Flutter environment with loadAppFonts() active, via flutter test --update-goldens --tags golden. - Review and cleanup gates: Inspect every changed PNG by eye, delete orphan goldens left by renamed tests, remove failures/ artifacts, and commit images as their own commit naming the cause. - Verification before merge: Re-run flutter test --tags golden twice without the flag to prove the new baselines reproduce, and confirm --update-goldens appears nowhere in CI. - Use Case: A design review raises a row height from 40 to 48 pixels and dozens of matchesGoldenFile tests go red. Follow the ritual to confirm the change is intentional, bless the full affected set in the pinned environment, and land the images as a standalone commit. ## Quick Start A deliberate visual change broke my golden tests — walk me through the sanctioned re-baselining ritual and verify the new baselines reproduce without the update flag.
