--- name: linkedin-post-engagers description: Turn the people who reacted to or commented on a LinkedIn post into a vetted HeyReach lead list. Scrapes engagers with Apify, enriches their companies, qualifies each person against your ICP with subagents, shows the list for approval, then loads approved leads into a new HeyReach list. Use for /linkedin-post-engagers ["List name"]. argument-hint: "[,...] [\"List name\"]" allowed-tools: Read, Write, Bash, WebFetch, WebSearch, Agent, AskUserQuestion --- # LinkedIn post engagers → HeyReach list Given one or more LinkedIn post URLs: collect everyone who reacted or commented, look up their companies, qualify each person against the ICP below, show the list for approval, and load the approved people into a fresh HeyReach list. The HeyReach campaign and its messages are set up in the HeyReach UI; this skill only builds and loads the list. Published by Kestrel GTM (kestrelgtm.com/skills/linkedin-post-engagers). ## Setup (once) Environment variables: - `APIFY_TOKEN`: an Apify API token (free plan works for a few posts a month) - `HEYREACH_API_KEY`: HeyReach → Settings → API Edit the ICP section below before the first run. It is the whole qualify bar. ## ICP (edit this) - Keep: **{{BUSINESS_MODEL}}** companies (e.g. B2B SaaS) × personas **{{PERSONAS}}** (e.g. founder, GTM leader, exec). - Drop: students, open-to-work, recruiters, individual contributors outside the personas, agencies or consultants selling what you sell, B2C. - **Drop competitors**: companies whose product overlaps yours: **{{COMPETITOR_CATEGORIES}}**. They pass every firmographic filter, so check explicitly. ## Apify calls (async: start, poll, fetch) Actors (harvestapi): `harvestapi~linkedin-post-comments`, `harvestapi~linkedin-post-reactions`, `harvestapi~linkedin-company`. ```bash # start a run RUN=$(curl -s -X POST "https://api.apify.com/v2/acts/harvestapi~linkedin-post-comments/runs?token=$APIFY_TOKEN" \ -H 'content-type: application/json' \ -d '{"posts":[""],"maxItems":0,"profileScraperMode":"main"}' | jq -r .data.id) # poll until SUCCEEDED curl -s "https://api.apify.com/v2/actor-runs/$RUN?token=$APIFY_TOKEN" | jq -r .data.status # fetch results DS=$(curl -s "https://api.apify.com/v2/actor-runs/$RUN?token=$APIFY_TOKEN" | jq -r .data.defaultDatasetId) curl -s "https://api.apify.com/v2/datasets/$DS/items?token=$APIFY_TOKEN&clean=true" > comments.json ``` Same pattern for reactions (same input) and companies (`{"companies":["", ...]}`). Always use the async runs endpoint: viral posts take longer than the synchronous endpoint allows. Output shapes vary; read the person from `actor`, `author` or `profile`, whichever exists. ## HeyReach calls (POST only, header `X-API-KEY`) ```bash H="https://api.heyreach.io/api/public" # create a list (name max 50 characters) curl -s -X POST "$H/list/CreateEmptyList" -H "X-API-KEY: $HEYREACH_API_KEY" -H 'content-type: application/json' \ -d '{"name":""}' # read the leads already in a list (paginate with offset until a page comes back short) curl -s -X POST "$H/list/GetLeadsFromList" -H "X-API-KEY: $HEYREACH_API_KEY" -H 'content-type: application/json' \ -d '{"listId":123,"offset":0,"limit":100}' # add leads (max 100 per request). The URL field is profileUrl, NOT linkedInUrl. curl -s -X POST "$H/list/AddLeadsToListV2" -H "X-API-KEY: $HEYREACH_API_KEY" -H 'content-type: application/json' \ -d '{"listId":123,"leads":[{"profileUrl":"https://www.linkedin.com/in/...","firstName":"","lastName":"","position":"","companyName":"","location":""}]}' ``` Check `addedLeadsCount` and `failedLeadsCount` in the response. `0/0/0` means the leads were not parsed at all (usually the wrong URL field). ## Steps 1. **Parse args.** Post URL(s) and an optional list name (default: `Engagement - - `, trimmed to 50 characters) or the id of an existing HeyReach list to add to. Check both env vars exist. 2. **Scrape engagers.** Comments and reactions for every post. Merge into one record per person by LinkedIn URL. Someone who commented and reacted is one person, kept as a comment (higher intent, and keep the comment text). 3. **Cheap noise cut.** Before any reasoning, drop open-to-work, students, interns, recruiters, company pages and titles clearly outside the personas. When unsure, keep. 4. **Enrich companies.** One company scrape per unique company URL (cache repeats): industry, about, specialties, headcount, domain. 5. **Qualify with subagents.** Batches of 15 to 25 people per subagent. Each returns, per person: `{ linkedin_url, persona, fits_business_model, is_competitor, confidence, reach_out, reasoning }`. Persona is judgment, not string matching. `reach_out` is true only for a matching persona at a matching company that is not a competitor. Web lookup only when company data is thin. Keep one output file per batch and join results by position, not by URL: agents sometimes rewrite URLs. 6. **Dedupe.** Drop anyone already loaded before, so overlapping posts do not re-add the same people. Check every reach-out and borderline person against `heyreach-loaded.csv` (every profile URL this skill has loaded before) and, when you pass the id of an existing HeyReach list, against the leads already in it (`list/GetLeadsFromList`). Compare normalized URLs (lowercase, no trailing slash or query string). Note who was dropped. 7. **Review gate.** Show two lists, commenters first: **Reach out** (confidence ≥ 0.75) and **Borderline** (0.60 to 0.75), each with name, title, company, persona, one-line reason and the comment if there is one. Add a tally: scraped → after noise cut → qualified → after dedupe → reach out. Ask for approval (AskUserQuestion). Never load anything without an explicit yes. 8. **Load.** Create the HeyReach list (or use the existing list id you passed) and add the approved leads in batches of 100. Append every added profile URL to `heyreach-loaded.csv`. Report the list name, id, added and failed counts, so it can be attached to a campaign in HeyReach. ## Notes - Pick posts whose engagers are your buyers, not your peers. A post by an influencer in your own field is mostly peers, agencies and competitors. - Stay within LinkedIn's limits: roughly 25 to 30 connection requests per day per sender. - Outreach should open on the topic of the post, never on the fact that they engaged with it.