Social Creator Scraper — daily LinkedIn + X discovery of partnership candidates

You are helping me build a Social Creator Scraper agent. It runs once a day, searches LinkedIn and X for creators posting about topics relevant to my product, filters for accounts with meaningful audiences (1k+ followers by default), enriches each with profile data + recent posts + a verified email, and saves new creators to my workspace database.

The creators it finds are partnership / collaboration / sponsorship candidates — people with audiences who already talk about the space I'm in. Pair this with the Social Creator Outreach Writer blueprint to actually pitch them.

REFERENCE DOCS (read these before writing code)
- agntdata API documentation: https://agnt.mintlify.app/apis/overview
- LinkedIn endpoints used: `Search_Posts`, `Get_Profile_data_using_LinkedIn_URL`
- X endpoint: `Search` (twitter154)
- agntdata orchestrator: `people_email_finder`, `people_email_verifier`

ABOUT MY MOTION
- Product: <YOUR PRODUCT>
- One-line description: <WHAT IT DOES>
- ICP: <WHO BUYS IT — role + company stage>
- Why creator partnerships make sense for me: <REASON — e.g. "my buyer follows these creators", "I want sponsored content from them", "I want them building on my product">

TOPIC TARGETS
List 4–8 topic terms your ideal creators post about. The agent uses these as LinkedIn + X keyword searches. Examples for GTM tooling:
- "AI SDR", "GTM Engineer", "outbound automation", "agentic GTM", "sales automation", "AI agents"

Be specific enough that the matches are creators-in-your-space, not generic marketers.

FILTERS
- Minimum follower count: <MIN_FOLLOWERS> (default 1000 — anything less rarely has enough audience to move pipeline)
- Maximum follower count: <MAX_FOLLOWERS> (default 1,000,000 — mega-creators don't reply to cold outreach)
- Language: <LANG> (default English)

WHAT TO BUILD
- A scheduled agent on agntdata that runs daily (claude-sonnet-4-6 — topic-relevance judgment benefits from the smarter model).
- Each run: search LinkedIn + X for the topic terms, walk top results, fetch profile data, filter by follower count, run an email-finder + verifier waterfall, save new creators to `social_creators`.
- Filters as gates: prune early. Skip a creator BEFORE paying for email enrichment if they fail follower / topic / language gates.
- Dedup: never re-insert a creator whose LinkedIn URL or X handle is already in the table.

DATABASE
`social_creators` is the source-of-truth table this blueprint stands up. Minimum columns:
- id (uuid PK), platform ('linkedin' | 'x'), profile_url, handle, full_name, headline, follower_count, recent_posts (jsonb), topic_match (text — which topic term surfaced them), email, email_verified, email_status, status ('new' | 'contacted' | 'replied' | 'partnered' | 'declined'), discovered_at, contacted_at.

DELIVERY
- One-line JSON summary per run: searches run, candidates found, after follower filter, after dedup, emails found, new creators saved.
- DO NOT push leads anywhere. The downstream outreach writer does that.

GUARDRAILS
- Topic relevance is the most important filter. If a creator showed up in your search but their profile + recent posts are about something unrelated, drop them. The LLM owns this judgment.
- Cap email-finder calls — never run the waterfall on a creator who failed the follower or topic gate.
- Dedup against `social_creators.profile_url` AND `social_creators.handle` before any expensive call.
- Cross-platform dedup: if you found the same person on LinkedIn AND X, merge into one row (whichever you found first; tag both URLs).

When you're ready, ask me the ABOUT MY MOTION, TOPIC TARGETS, and FILTERS blocks.