follower-scrapelisted

# Follower Scrape Collects the raw audience. It does not judge fit — that is stage 6 — and it never drops the label saying where a person came from. ## Pipeline contract **Stage 5 of 20.** Reads `04-audiences.csv` (rows with `selected=true`). Writes one file per audience to `05-audience-raw/<audience_type>-<name-slug>.csv`. Files live in `runs/<run>/`. Column definitions: [`/prospecting-system`](../prospecting-system/SKILL.md). Every output row gets `source_type` and `source_detail` added: | Collected how | `source_type` | |---|---| | follows a LinkedIn page | `follower` | | reacted to or commented on a post | `engager` | | member of a community | `member` | | registered for or attended an event | `attendee` | A follower is a weaker signal than an engager. Keep them distinct; stage 11 ranks on it. ## Route A — LinkedIn page followers (GetLeads) Works the same for competitor, tool, and event pages: each is a LinkedIn company page. 1. `lookup_company_linkedin_followers` with the page URL and the ICP filters (`company_sizes`, `seniorities`, `countries` from `02-client-profile.yaml`). Free. Billing is on the requested count; non-matching leads are refunded after delivery. 2. Show the human: follower count, price, and the run's remaining spend cap. **Ask how many followers to buy and which email receives the CSV.** Minimum order is 2,500 leads ($8.75); smaller pages are billed the minimum. 3. Add the cost to `spent_usd` in `state.json`. If it would pass the cap, stop and