Guessing a query's intent from its wording is an easy way to build the wrong page — the same noun without qualifiers can lead to a purchase in one case and a review article in another. The actual search results never lie about what Yandex considers the intent to be.
What it does
For up to about ten queries, the skill checks frequency to prioritize them, then looks at the real Yandex SERP for each significant one — articles, catalogs, product cards, aggregators or video — and classifies the intent from what's actually ranking: informational, commercial, navigational or mixed, noting the split for mixed cases. This matters most for queries that look unambiguous on paper but hold a genuinely mixed SERP in practice, where a single page type won't capture the whole demand.
What you get
A chat table: query → frequency → intent → the page type that should serve it, with a short reason based on the SERP makeup, plus warnings where the intent doesn't match what you expected — say, a commercial page was planned but the SERP is dominated by informational articles, which means a different format is needed rather than a rewrite of the existing page.