scrape-post-eg

name	scrape-post-eg
description	Scrape every person who liked or commented on a LinkedIn post, filter out companies and the post author, optionally filter by headline keywords, and write Name, Headline, LinkedIn URL, and Engagement type to a Google Sheet. Use when given a LinkedIn post URL. Requires Apify MCP and gws CLI.
argument-hint	[linkedin-post-url] [google-sheet-url]

LinkedIn Post Engagers → Google Sheet

Given a LinkedIn post URL and a Google Sheet URL, this skill scrapes every person who liked or commented, removes company accounts and the post author, deduplicates by name, applies optional ICP headline filters, and writes a clean sheet: Name | Headline | LinkedIn URL | Engagement.

---

Required Inputs

Input	How to extract	Example
Post URL	First argument or URL containing linkedin.com/posts/ or linkedin.com/feed/update/	https://www.linkedin.com/posts/username_slug-activity-123/
Sheet URL or ID	Second argument or URL containing docs.google.com/spreadsheets/d/. Extract the Sheet ID from between /d/ and the next /.	https://docs.google.com/spreadsheets/d/1Abc...XYZ/edit → Sheet ID: 1Abc...XYZ

If the user provides only a post URL, ask for the Google Sheet URL before proceeding.

---

Actual Apify Response Shapes (verified 2026-03-23)

Understanding these prevents field-path bugs. The two actors return different structures.

Reactions (harvestapi/linkedin-post-reactions):

{
  "reactionType": "LIKE",
  "actor": {
    "id": "ACoAAED1AnMB...",
    "name": "Hassan Shahzad",
    "linkedinUrl": "https://www.linkedin.com/in/ACoAAED1AnMB...",
    "position": "BSc (Hons.) Mathematics & Economics | Lums"
  }
}

• No actor.headline — use actor.position
• No actor.location — not returned in short mode
• No actor.type — can't distinguish person/company from fields alone
• URLs are always internal ID format (/in/ACoAAA...)
• Company pages use /company/name/ URLs — this is how you filter them

Comments (harvestapi/linkedin-post-comments):

{
  "commentary": "The actual comment text",
  "actor": {
    "id": "ACoAAFM90Y4B...",
    "type": "profile",
    "name": "Monika Grycz",
    "linkedinUrl": "https://www.linkedin.com/in/grycz-monika",
    "position": "outbound x content x personal brands",
    "author": false
  }
}

• No actor.headline — use actor.position
• No actor.location — not returned in short mode
• actor.type: "profile" or "company" — use this to filter companies
• actor.author: true for the post author's own comments/replies — use this to filter the author
• Company accounts have actor.id: null and URL like /company/name/posts
• Person URLs are vanity slugs (/in/firstname-lastname-123)

---

Filter Parameters (set before running)

Specify these at the start of the request, or apply defaults:

Parameter	Example	Behavior
headlineIncludeKeywords	["estimator", "operations", "project manager"]	Keep if headline contains any keyword (case-insensitive). Leave empty = keep all.
headlineExcludeKeywords	["recruiter", "student", "consultant"]	Drop if headline contains any keyword (case-insensitive). Applied after include filter.

If no filters are specified, skip Step 4 and write all cleaned records.

---

Step 1 & 2: Scrape Reactions and Comments (run in parallel)

Both actors are independent — call them at the same time to halve total runtime. Use async: false (the default) so each call blocks until results are ready.

Reactions

Call actor harvestapi/linkedin-post-reactions via Apify MCP (call-actor):

{
  "posts": ["<POST_URL>"],
  "maxItems": 1000,
  "profileScraperMode": "short"
}

From each result, extract:

• actor.name → Name
• actor.position → Headline
• actor.linkedinUrl → LinkedIn URL
• reactionType → map to friendly label (see table below)

Reaction type mapping:

Raw value	Sheet label
LIKE	Liked
PRAISE	Celebrated
EMPATHY	Supported
APPRECIATION	Loved
INTEREST	Insightful
ENTERTAINMENT	Funny
(anything else)	(write raw value as-is)

---

Comments

Call actor harvestapi/linkedin-post-comments via Apify MCP (call-actor):

{
  "posts": ["<POST_URL>"],
  "maxItems": 1000,
  "profileScraperMode": "short",
  "scrapeReplies": false
}

From each result, extract:

• actor.name → Name
• actor.position → Headline
• actor.linkedinUrl → LinkedIn URL
• actor.type → for company filtering
• actor.author → for author filtering
• Engagement label: "Commented"

---

Step 3: Clean, Deduplicate, and Merge

Apply in this order: remove junk → remove companies → remove author → deduplicate.

3a. Remove company accounts

From reactions: Drop any record where actor.linkedinUrl contains /company/. From comments: Drop any record where actor.type === "company".

Report: "Removed X company accounts."

3b. Remove the post author

Extract the author's name from the comments data — look for any record where actor.author === true. Store the author's normalized name.

If found, also drop any record from reactions whose actor.name.trim().toLowerCase() matches the author name.

If no author is detected in comments (e.g., author didn't comment), extract the author slug from the post URL itself (e.g., john-doe-12345abc → approximate name "john doe") and use that for name matching.

Report: "Removed post author: [name]."

3c. Deduplicate by name

Why name-based dedup: Reactions return internal ID URLs (/in/ACoAAA...) while comments return vanity slugs (/in/firstname-lastname-123). The same person gets two different URLs, so URL-based dedup misses overlaps.

Dedup key: actor.name.trim().toLowerCase(). Build a map keyed by this value.

For each person:

Appeared in	Engagement column
Reactions only	Reaction label (e.g., "Liked")
Comments only	"Commented"
Both	Reaction label + " + Commented" (e.g., "Liked + Commented")

Merging rules when name matches:

• URL: Keep the vanity URL (the one that does NOT contain ACoAAA). If both are vanity or both are IDs, keep whichever is non-null.
• Headline: Keep whichever actor.position is non-empty.

Missing name: If actor.name is blank or null, fall back to linkedinUrl as the dedup key. If both are missing, skip the record.

Same name, different people: Rare edge case. Acceptable to merge; the LinkedIn URL will resolve to one of them.

Report: "Deduplicated: X reactions + Y comments → Z unique people."

---

Step 4: Filter by ICP (apply if filter parameters were set)

Apply in order:

4a. Headline include filter (if headlineIncludeKeywords is non-empty):

• Keep the record if the headline contains any of the include keywords (case-insensitive substring match)
• If headline is blank and include keywords are set → skip

4b. Headline exclude filter (if headlineExcludeKeywords is non-empty):

• Drop the record if the headline contains any of the exclude keywords (case-insensitive)
• Applied after include filter — exclusions always win

After filtering, report: "Filtered Z → N records (M dropped by headline filter)."

---

Step 5: Write to Google Sheet

Use the Sheet ID extracted from the user-provided Google Sheet URL.

First, clear the sheet (prevents stale data from previous runs):

gws sheets spreadsheets values clear \
  --params '{"spreadsheetId": "<SHEET_ID>", "range": "Sheet1"}'

Then write headers + all rows in one call:

gws sheets spreadsheets values update \
  --params '{"spreadsheetId": "<SHEET_ID>", "range": "Sheet1!A1", "valueInputOption": "RAW"}' \
  --json '{"values": [
    ["Name", "Headline", "LinkedIn URL", "Engagement"],
    ["Jane Smith", "VP Ops at BuildCo", "https://www.linkedin.com/in/janesmith", "Liked"],
    ["Bob Lee", "Estimator at GC Inc", "https://www.linkedin.com/in/boblee", "Commented"]
  ]}'

Build the full values array from the filtered data before writing — one gws call writes everything.

For any dataset with special characters or 200+ rows: Write the JSON to a temp file, then use escaped double-quote --params (not single quotes) with $(cat) for --json:

gws sheets spreadsheets values update \
  --params "{\"spreadsheetId\": \"<SHEET_ID>\", \"range\": \"Sheet1!A1\", \"valueInputOption\": \"RAW\"}" \
  --json "$(cat /tmp/engagers.json)"

Why not single quotes for --params? When --json uses $(...) expansion, the shell can corrupt single-quoted --params. Escaped double quotes avoid this. See learnings.md for details.

---

Output Columns

Column	Source	Notes
A: Name	actor.name	Full name
B: Headline	actor.position	Current role/title (this field is called position in Apify, not headline)
C: LinkedIn URL	actor.linkedinUrl	Full profile URL — prefer vanity slug over internal ID
D: Engagement	Derived	"Liked", "Commented", "Liked + Commented", etc.

---

Scale Guidance

• Most posts: maxItems: 1000 covers everything
• Viral posts (5K+ reactions): increase maxItems to 5000 on both actors
• Reactions and comments are paginated internally by Apify — no manual pagination needed

---

Valid Post URL Formats

Both actors accept:

• https://www.linkedin.com/posts/username_slug-activity-XXXXXXX/
• https://www.linkedin.com/feed/update/urn:li:activity:XXXXXXX/
• https://www.linkedin.com/feed/update/urn:li:ugcPost:XXXXXXX/

Strip tracking params (?utm_source=...) before passing to actors — they work fine without them.

---

Notes

• No location data in profileScraperMode: "short". Country filtering requires switching to "full" mode (slower, more expensive). This skill uses "short" for speed.
• actor.headline does not exist — the actual field name from Apify is actor.position. This skill uses the correct field.
• Company pages sometimes react to and comment on posts. They are filtered out in Step 3a.
• The post author often reacts to their own post and replies to comments. They are filtered out in Step 3b.
• See learnings.md at the repo root for additional gotchas discovered during development.