Every page is assigned one primary intent before writing starts.
We check answer-first flow, section logic, and user decision points.
We verify key claims, remove weak assertions, and normalize terminology.
We reject mixed-language output and force human-readable localization quality.
We enforce cluster contracts so pages link up to hubs and down to supporting guides.
Updates are triggered by factual changes, user feedback, indexing issues, or editorial quality checks.