thalarch-visual-qalisted

# Thalarch Visual QA The final pixels are the source of truth for visual claims. ## 1. Derive a visual checklist Convert the user's visual requirements into checks such as: - required subject/content; - composition/framing; - exact text; - style/brand match; - palette; - dimensions/aspect ratio; - alpha/transparency; - crop/safe zones; - elements that must be absent; - invariants preserved from the baseline; - responsive/runtime states. Use `PASS`, `FAIL`, `UNVERIFIED`. ## 2. View the whole artifact first Before zooming into details, inspect: - hierarchy; - balance; - focal point; - crop; - readability; - contrast; - overall visual coherence; - whether it obviously violates the requested direction. For polished/professional work, also ask whether there is any obvious visual weakness that would benefit from one small targeted improvement: weak focal hierarchy, awkward spacing/crop, unnecessary clutter, or a brand/reference mismatch. Do not turn this into a style blacklist. Gradients, glow, symmetry, bokeh, dramatic lighting, 3D, minimalism, or decorative effects are valid when they work for the image. Then inspect details. ## 3. Mechanical image probes Use the bundled read-only scripts when relevant: ```text python scripts/image_probe.py <image> python scripts/image_compare.py <baseline> <candidate> --out <diff.png> ``` They can prove properties such as dimensions, format, alpha, and same-size pixel change statistics. They cannot decide whether the design is bea