Competitor Post Engagers

Competitor Post Engagers

Audience-capture prospecting against a competitor's content. The premise: someone who likes or comments on a competitor's LinkedIn post is in-market and category-aware — they're already evaluating this space. This skill converts that audience into a tiered lead list, tagging each lead with the competitor and post they engaged.

It runs the shared engager -> qualified-lead engine: discover posts -> extract engagers (cost gate) -> dedup -> enrich selected (cost gate) -> score + tier -> export.

When to use

• "Find leads engaging with [competitor]'s LinkedIn posts."
• "Who's commenting on [competitor]'s content — are any of them ICP?"
• Audience-capture prospecting against one or more competitor company/profile pages.

Quick start with example configs

Ship-ready ICP configs live in ${SKILL_DIR}/configs/ — copy one and tune it instead of authoring from scratch:

• configs/icp.b2b-saas-revops.json — RevOps / GTM buyers at mid-market B2B SaaS (heads/VPs of sales, RevOps, growth; 51–1000 employees; software/SaaS).
• configs/icp.devtools-eng-leaders.json — engineering decision-makers who buy developer tooling / platform / DevOps (CTO, VP/Head of Eng/Platform/DevOps; devtools keywords like kubernetes, ci/cd, observability).

cp ${SKILL_DIR}/configs/icp.b2b-saas-revops.json ${WORKSPACE}/icp.json   # then edit
# (default Apify actors are pre-wired — see "Known-good actors" below; no --actor needed)
python3 ${SKILL_DIR}/scripts/find_competitor_posts.py --profile-url "<competitor>" --estimate-only

Known-good actors

The Apify scripts ship sensible public-marketplace defaults, so the agent does not have to supply --actor. They are swappable — pass --actor <user~actor-name> (or set the env var) to use any actor whose input takes the same fields. Defaults assume the harvestapi LinkedIn actor family (popular, public, no LinkedIn cookie required); confirm current pricing on the actor's Apify Store page.

Swap example: --actor apimaestro~linkedin-post-reactions-comments-engagements (another public reactions actor) — output is normalized the same way regardless of source.

Setup

cd ${SKILL_DIR}/scripts && npm install && npx playwright install chromium   # only if you need the Playwright degrade

Python scripts are stdlib-only (no install). At least ONE extraction path must be configured — see Credentials / env.

Workflow

Step 0 — define your ICP + competitors (you, the agent)

Copy a shipped ICP — ${SKILL_DIR}/configs/icp.b2b-saas-revops.json or icp.devtools-eng-leaders.json (or the minimal ${SKILL_DIR}/scripts/icp.example.json) — to ${WORKSPACE}/icp.json and edit it for the target ICP (titles, seniorities, employee ranges, industries, keywords, exclusions). List the competitor LinkedIn company/profile URLs you want to mine.

Step 1 — discover recent competitor posts

python3 ${SKILL_DIR}/scripts/find_competitor_posts.py \
  --profile-url "https://linkedin.com/company/acme" \
  --competitor "Acme" --days-back 30 --top-n 3 \
  --estimate-only                       # prints projected Apify cost, no spend
# happy with the cost? re-run with --yes:
python3 ${SKILL_DIR}/scripts/find_competitor_posts.py \
  --profile-url "https://linkedin.com/company/acme" \
  --competitor "Acme" --days-back 30 --top-n 3 \
  --yes --output ${WORKSPACE}/acme_posts.json

Ranks the competitor's recent posts by engagement and keeps the top N (cost control — you only mine the posts with the biggest audiences). With no APIFY_API_TOKEN it degrades to printing a web-search query you run with your own search tool to find the post URLs by hand.

Step 2 — extract engagers (COST GATE)

# Estimate first (no spend, no --yes):
python3 ${SKILL_DIR}/scripts/extract_engagers.py \
  --source apify \
  --post-urls "https://linkedin.com/posts/...,https://linkedin.com/posts/..." \
  --competitor "Acme" --estimate-only

# Confirm the spend with --yes:
python3 ${SKILL_DIR}/scripts/extract_engagers.py \
  --source apify \
  --post-urls "https://linkedin.com/posts/...,https://linkedin.com/posts/..." \
  --competitor "Acme" --yes \
  --output ${WORKSPACE}/acme_engagers.json

extract_engagers.py pulls reactors + commenters and tags every row with the competitor and source post. Multi-source fallback (use --source):

• apify (default, APIFY_API_TOKEN) — managed actor with async run + poll and the cost-confirm gate (--estimate-only / --yes). Configure the actor with --actor (default a harvestapi-style reactions/comments actor).
• phantombuster (PHANTOMBUSTER_API_KEY + LinkedIn cookie on the phantom, pass --engagers-agent-id).
• playwright (LI_AT cookie) — keyless local-browser degrade, lowest volume.

Output rows: {name, headline, profile_url, engagement_type, comment_text?, post_url, competitor, source}.

Step 3 — dedup across runs (credit saver)

python3 ${SKILL_DIR}/scripts/dedup_history.py \
  --input ${WORKSPACE}/acme_engagers.json \
  --history ${WORKSPACE}/engager_history.csv \
  --output ${WORKSPACE}/new_engagers.json

Drops anyone already processed in a prior run (workspace CSV; mirrors to Supabase if SUPABASE_URL/SUPABASE_KEY are set). Run this BEFORE enrichment so Apollo credits are only spent on net-new people. Use --dry-run to preview the net-new count.

Step 4 — select, then enrich the survivors (COST GATE)

Read new_engagers.json and select the ones worth enriching (the engine scores everyone, but you control who gets Apollo credits — e.g. skip obvious non-fits by headline). Then:

python3 ${SKILL_DIR}/scripts/enrich_apollo.py \
  --input ${WORKSPACE}/new_engagers.json \
  --limit 50 --reveal-email \
  --output ${WORKSPACE}/enriched.json

Resolves {title, seniority, company, company_domain, company_size, industry, email?} via a two-phase Apollo call: (A) if a row has a company name but no domain, resolve the Apollo organization to a primary domain (cached per company), then (B) People Match keyed by name + domain for a higher hit rate (engagers usually arrive name + headline only). With no APOLLO_API_KEY it passes through profile-only (no spend, no error) so the scorer still runs. Optional email fallback: set DROPCONTACT_API_KEY to guess an email from domain + name when Apollo returns none (used only with --reveal-email).

Step 5 — score + tier (deterministic)

python3 ${SKILL_DIR}/scripts/score_icp.py \
  --input ${WORKSPACE}/enriched.json \
  --icp ${WORKSPACE}/icp.json \
  --output ${WORKSPACE}/scored.json

Computes a 0-100 score = fit_weight*FIT + intent_weight*INTENT (defaults 0.6/0.4) and an A/B/C tier. FIT = firmographic match to your ICP; INTENT = engagement-type weight (a comment >> a like) x post recency. Every lead carries a scoring_breakdown for audit. Tune --tier-a / --tier-b / --half-life-days / weights as needed.

Step 6 — final review + export (you, the agent)

Read scored.json. Do the qualitative pass the scorer can't: read comment_text (is the comment a buying signal or a throwaway emoji?), sanity-check tier-A leads, demote false positives. Then export:

python3 ${SKILL_DIR}/scripts/export_csv.py \
  --input ${WORKSPACE}/scored.json --min-tier B \
  --output ${WORKSPACE}/competitor_engagers.csv

Hand the CSV to the user, import into contact-cache, push to a channel, or persist to Airtable/Supabase.

Outputs

• *_posts.json — top competitor posts by engagement.
• *_engagers.json — raw reactors/commenters, tagged with competitor + source post.
• new_engagers.json — net-new engagers after cross-run dedup.
• enriched.json — net-new engagers with firmographics (+ email if revealed).
• scored.json — every lead with icp_score, tier, and scoring_breakdown.
• competitor_engagers.csv — flat export, optionally tier-filtered.

Credentials / env

A LinkedIn engager source is the hard gate — provide at least one of:

• APIFY_API_TOKEN (optional but the primary path) — managed actors for both post discovery and engager extraction, with the cost-confirm gate.
• PHANTOMBUSTER_API_KEY (+ LinkedIn session cookie on the phantom; pass --engagers-agent-id) — --source phantombuster.
• LI_AT (LinkedIn li_at cookie) — keyless --source playwright degrade.

Optional enrichment / state:

• APOLLO_API_KEY — two-phase Apollo enrichment (org-resolve + People Match); without it, profile-only pass-through.
• MILLIONVERIFIER_API_KEY — enrich_apollo.py --verify email deliverability; without it, a keyless syntax + MX-record check.
• DROPCONTACT_API_KEY — optional email fallback when Apollo reveals no email.
• SUPABASE_URL / SUPABASE_KEY — mirror cross-run dedup history to Supabase (degrades to the CSV ledger).

Notes & edge cases

• Engagers of a competitor are warm, category-aware leads — they're evaluating the category right now. Each lead is tagged with which competitor + which post they engaged so outreach can reference it.
• Cost discipline: discovery and extraction both refuse to spend without --yes (--estimate-only prints the projection first); dedup runs before enrichment; you select who gets enriched. Never extract one call per post blindly — discover, rank, mine the top N.
• Intent > volume: a thoughtful comment outscores a passive like; a fresh post outscores a stale one. That's the engagement_type x recency intent half of the score.
• Apify/PhantomBuster may cap reactions/comments on high-virality posts — note partial coverage. Pair scraping with a proxy + throttle; LinkedIn anti-bot is aggressive at engager-list depth.
• The Playwright degrade is best-effort DOM scraping — expect lower coverage and missing headlines/locations; rely on Apollo enrichment to backfill firmographics.