Skip to content

Setup — "Best Posts" Automation (You + Influencers)

Adapts Ruben Hassid's method (see Ruben Hassid - Train Claude on Your Best LinkedIn Posts) into a repeatable pipeline that surfaces your best posts and the best posts of the influencers we already track in ../influencers.md.

Goal: two engines that share one pipeline — 1. You-viral — drafts new posts in your proven voice, from your own top performers. 2. Swipe-file — mines what's working for the adviser/adviser-tech influencers, so you can borrow structures and angles (not copy) and comment/engage with intent.

The Pipeline (one flow, two audiences)

Apify export (CSV)  →  Claude analysis (report + SOP)  →  Claude skill
   (per profile)          (what works & why)               (drafts / swipe file)

A. Your best posts → dean-viral skill

  1. Extract: Apify LinkedIn posts actor → your profile URL → export CSV. (Ideally 30+ posts; if thinner, still useful but treat findings as directional.)
  2. Analyse: paste the analysis prompt below, attach the CSV.
  3. Skill: /skill-creator → build dean-viral from the report. Reuse across chats.

B. Influencers' best posts → swipe-file skill

Run the same Apify export for each priority influencer, then have Claude analyse across all of them for transferable patterns — hooks, formats, cadence, post structures — not topics to copy. Prioritise the highest-leverage voices from ../influencers.md:

Tier Who Why first
Analyse first Ian McKenna, Abraham Okusanya, Michael Kitces, Ben Henry-Moreland They shape how the profession judges tools; highest reach into our ICP.
Then Warren Ingram, Craig Iskowitz, David Ferguson, Nicky Sevel Strong adviser / advicetech reach across UK + SA.

Batch cost: Apify runs are ~$2–5 each; the free $5 credit covers your first profile. Budget a few dollars to sweep the "analyse first" tier in one sitting.

Prompt 1 — Analysis (paste with the CSV attached)

I'm attaching an Apify "LinkedIn profile posts" export. Treat this as a serious analytics
deliverable, not a summary.

For each post: compute engagement = likes + comments + shares. Decode the post timestamp from
the activity ID. Classify format (text / image / video / carousel / article) and the hook/angle
of the opening lines. For any visual post, OPEN THE MEDIA and actually look at it — do not judge
from the caption alone.

Then produce TWO deliverables:
1. A written report: top performers with links, patterns by format and by hook, cadence vs.
   engagement, and clear outliers (what over/under-performed and a hypothesis why).
2. An SOP: a Stop / Continue / Start list that someone could follow to replicate the wins.

Be specific — cite exact posts, numbers, and links. No vague advice.

For the influencer batch, append: 

This is an influencer I want to LEARN FROM, not imitate. Focus on transferable STRUCTURE:
hook formulas, post formats, cadence, and how they open/close. Flag anything that only works
because of who they are (name recognition, audience size) vs. what any credible voice could
reuse.

Prompt 2 — Build the skill (same chat, after the report)

/skill-creator Create a skill from what matters in the report above.

Name it "dean-viral" (for the influencer version, name it "linkedin-swipe-file").

The skill should:
- Embed the concrete findings (winning hooks, formats, cadence) as its playbook.
- Ask me clarifying questions first (topic, goal, format) via AskUserQuestion.
- Then generate 3-5 post options, each with: hook, full caption, and media direction.
- Keep everything consistent with the historical top performers in the report.

How the two skills work together

  • Draft with dean-viral (your voice, your proven patterns).
  • Pressure-test structure/angle against linkedin-swipe-file (what's landing for the people our ICP already follows).
  • Feed engagement outcomes back in periodically (re-export, re-analyse) so both skills stay current — this is iteration, which is exactly what the method is good at.

Cadence & re-runs

  • Monthly / quarterly: re-export your profile, re-run Prompt 1, refresh dean-viral.
  • Ad hoc: when an influencer post pops, screenshot it and run Step 4 ("capture the recipe") from the source article to extract a one-off template.

Guardrails

  • Public data + official tools only (Apify), no automation on accounts — as the source frames it.
  • Swipe structure, not substance — borrowing an influencer's hook shape is fine; lifting their take is not, and it won't sound like you.
  • The method assumes a real cadence and enough history; thin data = directional, not gospel.