prospect-posts

Prospect Posts

You scrape the most recent LinkedIn posts of one or more profiles via Apify and scan them for a specific theme the user cares about (e.g. "AI-first GTM", "hiring pain", "pivoting to enterprise"). Output is a structured report showing which profiles mentioned the theme, with quoted excerpts and post links.

This is research for prospect/account intelligence - read-only, multi-profile.

How to invoke

The user says something like:

• "pull the last 20 posts from [profile URL] and look for mentions of [theme]"
• "scan these three founders' LinkedIn for talk of [topic]"
• "has [prospect] posted about [theme]?"

Required inputs:

1. Profile URL(s) - one or more LinkedIn profile URLs
2. Theme - what to look for. Can be a topic, belief, pain point, or signal

Optional:

• Count - posts per profile (default 20)
• Output path - where to write the report. Default derived from theme + date (see Step 4)

If either profile URL or theme is missing, ask the user before running.

Prerequisites

• APIFY_API_TOKEN in .env
• requests and python-dotenv installed

Process

Step 1: Prepare

1. Confirm APIFY_API_TOKEN is set. If missing, tell the user to add it.
2. Pick the output directory:
   • Single profile that maps to an existing per-prospect folder (e.g. prospects/{slug}/): save there
   • Otherwise: prospects/_scans/ (default)
3. Derive a filename slug from the theme (lowercase, hyphens, no punctuation) and today's date.
   • JSON path: {output_dir}/{date}-{theme-slug}.json
   • Report path: {output_dir}/{date}-{theme-slug}.md
4. Create prospects/_scans/ if it doesn't exist.

Step 2: Fetch posts

Run the scraper. Repeat --profile-url for each profile:

python3 scripts/prospect_posts.py \
  --profile-url "<url-or-username-1>" \
  --profile-url "<url-or-username-2>" \
  --count 20 \
  --output-path "<json-path>"

The script:

• Uses the apimaestro/linkedin-profile-posts actor (no LinkedIn cookies needed, $0.005/post)
• Starts one actor run per profile in parallel, then polls until all complete
• Accepts either a full URL (https://www.linkedin.com/in/foo/) or a bare username (foo)
• Uses the actor's total_posts input to auto-paginate to the requested count
• Writes structured JSON with {profiles: [{input, username, profile_url, name, headline, status, posts: [{date, url, type, text, engagement}]}]}
• Includes reshared-post text inline with a [Reshared from X] prefix so theme matching sees it
• If a run fails (FAILED/ABORTED/TIMED-OUT), that profile appears in the output with status set and an empty posts array - surface this to the user