Niche Discovery Skill

Purpose

Pick each additional channel's niche from real, current data — not from a
generic "good niche ideas" list. Every niche recommended here must trace
back to an actual search result.

Process (per candidate niche)

1. Broad category sweep. Search YouTube (via trend-research) across
   candidate categories not already covered by existing channels — pull
   real view counts, upload frequency, and recency for top channels/videos
   in each.
2. Durability check (separate pass, don't skip this). Search the web
   for whether this category is a short-lived spike (one viral event) or
   has a track record of sustained interest — look for multi-month/year
   search or view trend signals, not just this week's numbers. A niche
   that's only hot because of a single recent event should be flagged as
   lower-durability, not silently excluded or included.
3. Practitioner/community check. Search what other automation/niche
   YouTubers say about this category's monetization and audience
   durability — this is a real signal, cite it, don't guess at it.
4. Style fit. Based on what the actual top-performing videos in that
   niche look like (checked via search, not assumed), recommend one of the
   three existing styles:
   • fact/list/stat-heavy content → minimal
   • narrative, personal, or emotionally-driven content → the researched
     third style (see remotion-render skill — determined by research, not
     fixed in advance)
   • process/explainer/how-it-works content → motion-graphics
5. De-duplication. Check the new niche isn't a near-duplicate of an
   already-configured channel's niche (compare against config/channels.json).

Output (per niche, before creating any channel config)

• Niche name
• Evidence it's currently performing well (with real numbers/sources)
• Evidence for or against durability (explicit, not assumed)
• Recommended style + why
• Any practitioner/community notes found

Log this to a niche-candidates.md (or similar) file in the repo as each
niche is confirmed, so there's a durable record of why each channel's
niche was chosen — don't only put this in chat output.

Rules

• Never propose a niche without at least one real search backing the
  "currently performing well" claim and a separate real search backing
  the durability claim — these are two different questions, don't
  conflate them into one search.
• Never fabricate view counts, growth percentages, or trend language —
  if a number wasn't returned by an actual API/search call, don't state a
  specific number at all; describe it qualitatively instead ("multiple top
  videos in the last month" rather than a made-up percentage).
• If fewer than 45 niches turn up strong evidence, report that back rather
  than padding the list with weak candidates — quality over hitting the
  exact number.