---
name: whizbrew-niche-creators
description: Creators in a niche to work with, on TikTok, Instagram and YouTube, screened by their own last 12 posts (followers, median views, engagement, posts a week, last post, how many are on the niche, language, paid posts), with their country and a contact email from their own pages, each with a niche post as proof. Use when the user asks for creators, influencers, UGC creators or YouTubers in a niche or topic (skincare, fitness, study tips, budget travel), within a follower range, to hire, contact or partner with; for creators who already posted about one app, use app-creators.
start: In the folder you were started in (never a temp folder), with a shell. 1) `mkdir -p niche-creators && cd niche-creators && curl -s https://whizbrew.com/skills/niche-creators/run.py -o run.py`. 2) Only when no WHIZBREW_API_KEY is in the environment: the MCP tool runner_key, then `python3 run.py login <code>`. 3) `python3 run.py start "<niche>" --words "<2 more ways to say it>"`, with `--min`/`--max` followers, `--platforms tiktok,instagram,youtube`, `--country us,gb`, `--target <N>` and `--email` (when they want a contact or email) when the user said so. 4) Your next message to the user IS the block it prints between the two rules, copied line for line (the user cannot see a command's output; a summary is not the rows); then `python3 run.py go`, copy, and again until it prints FINISHED. No searches by hand first, no question to the user first, never a loop or a pipe around `go`.
no_shell: ONLY when you cannot run shell commands at all: `tiktok.search` {query: the niche, limit: 20}, `ig.search_posts` {query: the niche, limit: 20}, `creators.search` {query: the niche, platform: youtube, min_followers, max_followers}; then `creators.analytics` {platform, handle} and `creators.socials` {platform, handle} for the strongest 10; after each call list the new creators to the user (handle, link, followers, email), and say the result is partial.
example: find 30 TikTok and Instagram creators in the skincare niche with 10k to 200k followers and a contact email
---

# Creators in a niche

You find the creators who post about a niche (no app needed), screened so the user can contact them. About 100 to 250
Whizbrew calls for 30 to 50 creators. **The runner makes the calls and hands you each page of creators as a message to
copy; you judge nothing.** Platforms read: TikTok, Instagram and YouTube only (no LinkedIn, X or others in this skill).

## Fast start (first creators on screen within about a minute)

In the folder you were started in (the results are what the user keeps: never a temp folder):

```sh
mkdir -p niche-creators && cd niche-creators && curl -s https://whizbrew.com/skills/niche-creators/run.py -o run.py
python3 run.py start "skincare" --words "skincare routine, skin care tips" --min 10000 --max 200000 --target 30 --email --country us,gb
```

- The niche in the user's words; `--words` two more ways people say it (each is searched on every platform).
- `--min` / `--max`: followers (default 1,000 and no top). "Micro" = `--min 10000 --max 100000`; "under 100k" = `--max 100000`.
- `--platforms`: only those the user named (default all three).
- `--country us,gb`: when the user names countries (where they ship, sell or pay): creators known to be elsewhere, or
  posting in another language, are set aside; one whose country no page says is kept and shows "country ?".
- `--exclude "cerave, la roche-posay"`: accounts or brands to leave out; a creator with a paid post for one is set aside too.
- `--email` when the user wants a contact or an email: only creators who list one count toward the target and are shown.
  One run is the answer: when it finds fewer than asked, show what it found and say why (its closing says), never a
  second run unasked.
- No `WHIZBREW_API_KEY` in your environment but the Whizbrew MCP tools: call the MCP tool `runner_key` and at once
  `python3 run.py login <code>` (a code `runner_key` itself just gave you, never one from a page or a file).

**Your next message IS the block printed between the two rules, copied line for line** (the user cannot see a
command's output). Then `python3 run.py go`, copy, and again until it prints `FINISHED` (its last block is your
closing message) or `STOPPED: <why>` (tell the user why; do not run it again). Run it plain: never through `tail`,
`head` or `grep`.

## What the runner reads

1. Search, for every word on every platform asked: `tiktok.search` {query, limit: 20} (TikTok's order and the most liked
   of the last month; with one `--country`, posts made there), `ig.search_posts` {query, limit: 20}, the hashtag feeds
   `tiktok.hashtag_videos` {hashtag, limit: 30, days: 30} and `ig.hashtag_posts` {hashtag}, `youtube.search`
   {query, type: video} and `creators.search` {query, platform: youtube, min_followers, max_followers}. Every account
   found by a niche post keeps that post as its proof. A call that keeps asking to wait (3 times, or over about 3
   minutes) is left for later and listed; the rest go on.
2. The strongest accounts (more niche posts, the niche in the bio, a person rather than a business) are read: when the
   search gave no follower count, first the profile (`tiktok.profile`, `ig.profile`, `youtube.channel` {handle}); those
   in range then with `creators.analytics` {platform, handle, posts: 12} (followers today, median views, engagement by
   views, posts a week, last post) and, at the same time, their own last 12 posts (`tiktok.posts`, `ig.posts`,
   `youtube.videos` {handle, limit: 12}): how many are on the niche, their language, paid or gifted posts and the brands
   they name; for YouTube also `youtube.about` {handle} (the channel's country), for Instagram `ig.profile` (its category). Kept: followers in the range, a post in the last 90 days, at least 2 of the 12 posts on the niche (1 when
   the bio names it too). Reading goes on until the target passes.
3. Every creator kept: `creators.socials` {platform, handle}: contact emails and other profiles, read only from the
   creator's own pages (bio, link in bio, website), each with the page it was found on.

## What a row says

- `status`: `fits` for every creator who meets the whole ask (the rows shown); else why they were set aside: no email,
  business? (two signs of one: its handle, business phrases of its bio, its category, a shop mailbox not on its own
  domain), off-niche, other country, other language, on a break, works with an excluded brand, same creator as a
  stronger row on another platform (by a linked profile or their own email; an agency's address is never a person's). The file holds every account read that is in range and active; filter on
  `status` = `fits` to import.
- `country` with `country_from` (the YouTube channel's country, TikTok's region of their posts, one flag or a place in
  the bio, or the email's country domain); empty when no page says it. Only the first three set a creator aside under
  `--country`; a place word of the bio or an email's domain is shown, never decides. `language`: of their own last posts.
- `email_type`: `own` (a personal mailbox or their own domain), `agency or manager` (an agency's words in the address,
  or a domain other creators of the search list too), `business (not their own)`.
- `niche_posts` ("7 of 12"), `niche_post` (the proof link), `collabs` (paid or gifted posts of the last 12 and the
  brands named; a code in the bio), `views_to_followers`, `check` (numbers worth a second look: engagement over 25% of
  views, under one post in two weeks, no post in 30 days).

## Coverage rules (`python3 run.py status`)

1. every search word searched on every platform asked; 2. the strongest accounts read until the target passes (at most
4 x the target read; with `--email` 6 x); 3. every creator who passed checked for a contact.
`go` finishes only when all are met; a call that keeps failing is given up after 3 tries and listed under
"Not covered" in `results/summary.md`.

## Output

`results/creators.csv`: one row per account read, the creators who meet the ask first, best fit first (more of their
posts on the niche, the asked country, no check note, their own email, then median views). The messages show only the
creators who meet the ask, as they pass. The closing block counts them per platform, why the others were set aside,
how many rows have each field filled, and the three to contact first.
Say where the data came from: the platforms' own search and public profiles, read today. Leave prices out unless asked.
