page-decomposition

Page decomposition analyzes a single section's content to identify distinct sequences and natural breaking points. Provide the section's visual description, screenshot, and cleaned HTML; the skill outputs neutral descriptions of structure (e.g., 'two images side-by-side,' 'grid of 8 items with icons') without assigning block names, enabling objective authoring decisions.

Neutral description captures what content exists and how it's arranged without predetermining its block type. This prevents bias during content modeling—you describe 'large centered heading, paragraph, two buttons' first, then decide the block structure afterward based on that objective analysis.

Apply page decomposition early during page import, analyzing each section individually to map its internal components before authoring decisions. It serves as the foundation for downstream content modeling and block assignment, ensuring structured, unbiased analysis of every section's layout and content flow.

The skill requires three inputs: the section's visual description and boundaries, a screenshot showing the section's layout, and cleaned HTML of the section content. These combined inputs enable precise identification of content sequences and breaking points for accurate structural representation.

Yes. Page decomposition is designed to pinpoint natural breaking points between different content types and patterns within a single section, identifying where one sequence ends and another begins—essential for complex sections mixing text, images, grids, and interactive elements.

The skill outputs neutral sequence descriptions detailing each content component, its arrangement, and identified breaking points. For example: 'Hero section contains large centered heading, descriptive paragraph, two call-to-action buttons, followed by two large images displayed side-by-side.'