VET Investigation

Read VET overlay images and identify structural discrepancies between two pages. This skill produces a list of issues — it does not measure computed styles or produce annotations.

Inputs required

• Paths to the images produced by vet-generator:
  • 2 original screenshots (standard + dev)
  • 2 VET overlay screenshots (standard + dev)
  • 2 diff images (side-by-side + pixel diff)
• Path to the VET info JSON (colour-to-semantic-role mapping)
• Standard page URL and dev page URL (for context)

Cropping for detail

VET screenshots are typically large (e.g. 2400×1636). Small elements like nav tabs or indicators are hard to distinguish at full size. Use the crop.mjs script to zoom into any region of interest before drawing conclusions: