ozerohax/testing-visual-regression

<input_requirements> <required>List of pages/components for visual coverage</required> <required>Baseline version and update rules</required> <required>Browsers and viewports</required> <optional>Dynamic areas to mask</optional> <optional>Diff sensitivity threshold</optional> <optional>Theme/locale requirements</optional> </input_requirements>

<preparation> <steps> <step>Stabilize data and state (fixtures/seed)</step> <step>Disable animations and auto-updates when needed</step> <step>Fix fonts, theme, and locale</step> <step>Fix viewport size and scale</step> </steps> </preparation>

<execution_rules> <rule importance="critical">Screenshots are reproducible and deterministic</rule> <rule importance="high">Comparison is done against the correct baseline</rule> <rule importance="high">Dynamic areas are masked or stabilized</rule> <rule importance="high">Changes are confirmed by a human (review)</rule> <rule importance="medium">The tool and comparison version are recorded</rule> </execution_rules>

<coverage> <focus> <item>Critical screens and user flows</item> <item>High-variance UI components</item> <item>Tables, forms, charts, modals</item> <item>Different viewports and themes</item> </focus> </coverage>

<quality_rules> <rule importance="critical">Expected output and baseline match by version</rule> <rule importance="high">Diff threshold is documented</rule> <rule importance="high">Results include artifact links</rule> <rule importance="medium">Reasons for baseline updates are recorded</rule> </quality_rules>

<do_not> <item importance="critical">Do not update baseline without reviewing changes</item> <item importance="high">Do not compare screenshots with unstable data</item> <item importance="high">Do not mix different locales/themes in one baseline</item> </do_not>

<example_checks> <check>Compare a product card across 3 viewports with fixed data</check> <check>Verify visual changes in a modal after an update</check> </example_checks>