Video SEO Optimiser

YouTube SEO is a dual-optimisation problem. Most creators do one half and assume the other will follow — either keyword-stuffing every field and getting indexed but never clicked, or writing punchy titles that nobody can find. This skill works both sides. Primary keyword goes in the title front-half and the first sentence of the description; clickability gets a differentiator like year, scope, or level; a parenthetical curiosity hook earns the click when the keyword has done the search work.

Titles get generated in three flavours — search-optimised, curiosity-optimised, and a hybrid recommendation — all under 60 characters because mobile truncates the rest. Descriptions are structured by zone: zone 1 above the fold (the only part most viewers see), zone 2 chapter timestamps with keyword-rich titles that rank independently in search, zone 3 the 150-300 word natural-language body where secondary and LSI keywords sit without stuffing, and zone 4 the links, hashtags, and related-video pointers that keep viewers in your content. Tags follow a tiered order — exact primary first, secondary, broad topic, brand — capped at 15-20 tags and 500 characters.

Chapter markers serve three jobs at once. They rank independently in search. They cut bounce because viewers can skip to the relevant bit. And Google shows them in search results, raising CTR. The skill enforces the rules — minimum three chapters, maximum ten, each at least ten seconds, first chapter must start at 0:00, titles specific and keyword-aware.

Where this earns its keep is the diagnostic work on existing libraries. High impressions + low CTR is a title or thumbnail problem; high CTR + low impressions is an SEO problem; good retention + low overall views is a distribution problem needing better metadata. Updating ten existing videos often outperforms publishing ten new ones, and the skill produces a priority list when the user has a back catalogue. It also matches keyword difficulty to channel authority — a 2,000-sub channel shouldn't fight for "how to lose weight" when there's a long-tail variant they can actually win. The output ends with a search-positioning note that names the differentiation angle and the expected competition level honestly.