Review the yard's visual states

Run the isolated fixtures with Chrome installed:

node scripts/visual-audit.mjs /tmp/playpets-visual-review

Open index.html in the output folder. The script serves repository files on an ephemeral loopback port, loads the real content, scenery, sprite, view and game modules, and substitutes the native bridge with an isolated fixture. It does not read or change the app's saved yard.

The gallery covers these fixture groups; its generated index reports the exact capture count:

• 48 season, time and weather combinations;
• five seasonal events;
• four collection books with discovered specimens in both system appearances;
• 24 studio combinations (four times of day, three weathers, light and dark appearance), plus two studio mini sizes, a wide strip and a portrait window;
• four window sizes, including the narrow mini window and a wide strip;
• game instructions, wrong guesses, discovery, tag transfer, formation, active laps and completion;
• reduced-motion game states, a pet reaction and evening rain.

Game fixtures assert the expected phase before capturing it. Fixed-step travel makes the fixtures repeatable; temporary poses are drawn against their real clock. Screenshots establish how those states compose, rather than frame-by-frame animation smoothness. Native end-to-end checks cover the app bridge and real interactions separately.

sound-report.json contains 48 offline renders at maximum volume with overlapping cues, birds, water and cricket sounds. Seeded noise keeps the stress scenario repeatable. The script rejects non-finite samples and clipping. This checks signal integrity, not whether someone finds a sound pleasant.

Review readability, clipped figures, coherent lighting, layered depth, visible progress and completion, and cleanup after replay. Compare changed art with the previous version before keeping it. Ordinary interface icons can remain vectors; generated art is useful when material, shape or scenery detail needs it.

After changing behaviour, run its focused regression checks, then npm test, npm run lint and npm run test:e2e. Run npm run site:build for docs or website changes. A successful fixture run does not replace first-use testing with people or validation on the other desktop platforms.