← all playbooks
Signals

Turn a LinkedIn connections export into a scored prospect database

Your entire LinkedIn network sorted into people who match your ideal client and people who do not, with contact details, plus a weekly shortlist of ten targets worth a personal message.

use when
you have thousands of first-degree connections and no idea which of them are actually buyers

The prompt

Paste it into Claude Code or the Claude desktop app with Hyreflow connected. The first line loads the recruiting skill, so your agent reads the play, asks before it spends anything, and hands the work back to you.

paste this into Claude
/hyreflow-recruit

Here is my LinkedIn connections export: <PATH_TO_Connections.csv>

Sort my network out for me.

1. Prefilter on title and function so we are not enriching students and
   people who left the industry.
2. Enrich what is left so each row has a company, a title and contact
   details where they exist.
3. Screen each person against this definition of a prospect:
   <PASTE YOUR PROSPECT DEFINITION>
   Give me a straight in or out with a reason. Do not give me a score out
   of a hundred, I will only start arguing with the number.
4. Save the result so I can query it later.

Then, separately, build me a weekly shortlist:
- Exclude anyone I have contacted in the last 90 days.
- Rank on who has been active recently.
- Take the top 10, research each one for anything worth mentioning, and
  draft me a connection note under 300 characters in my voice.

I will send them myself. Do not automate anything on LinkedIn.

Replace every <PLACEHOLDER> with your own detail. Everything else can stay as written.

What you need first

  • Your Connections.csv export from LinkedIn
  • A written description of who counts as a prospect for you
  • A Hyreflow workspace with credits
  • Somewhere to keep the result, such as your own spreadsheet or database

Tools it can reach for

The agent picks per step from what your workspace has. Nothing here is required by name.

What happens when you run it

Free steps are marked free. Anything that spends credits is marked, and the agent asks before the first paid run of any size.

  1. 1

    Read the export

    free

    Your Connections.csv is parsed as it comes out of LinkedIn, including the rows where the name encoding is mangled or the email is withheld, which is most of them.

  2. 2

    Prefilter before spending

    free

    Titles and functions that could not possibly be prospects are removed first. Enriching your entire network is the expensive mistake this step exists to prevent.

  3. 3

    Enrich what is left

    credits

    Current company, title and contact details, matched on the LinkedIn profile where possible and on name plus company otherwise.

  4. 4

    Screen each one, in or out

    credits

    A binary decision against your written definition, with the reasoning attached. Deliberately not a score: a number invites you to move the threshold until the answer flatters you.

  5. 5

    Store it where you can query it

    free

    The scored network lands in your own spreadsheet or database, so it becomes an asset you can filter next month rather than a one-off report.

  6. 6

    Suppress anyone contacted recently

    free

    First step of every weekly sweep, before anything is ranked. Contacting someone twice in a month is how a warm network becomes a cold one.

  7. 7

    Rank on recent activity and take ten

    credits

    People who have been visibly active are more likely to respond. The top ten are researched for something specific and current worth referencing.

  8. 8

    Draft the notes and stop

    credits

    Ten connection notes under the character limit, in your voice, staged for you. You send them by hand, up to twenty or twenty-five a day. Nothing is automated on LinkedIn, ever.

The most underused asset in a recruitment business

A consultant with ten years in a market has a few thousand connections. Somewhere in there are the hiring managers, the operations directors and the founders who make up their entire addressable market. They are first-degree connections, which means the message gets delivered and the name is familiar.

Almost nobody works it, because the export is a wall of names with no company, no title worth trusting and no way to tell a buyer from a former colleague from someone who added them at a conference in 2019.

Two things, built once

The database. Every connection screened in or out against a written definition of a prospect, with the reason. Built once, queried forever. This is the part that turns a contact list into something you can filter by industry, by seniority or by geography on the day you need it.

The weekly sweep. Ten people a week, chosen because they have been active, researched for something specific and current, with a draft note waiting. That cadence is deliberately small. It is what a person can sustain and what stays personal.

Why binary, and why the write-up matters

The screening decision is in or out, never a number. The reason is written down. That means two things: you can audit the decisions, and if a large majority of your network is coming back as a prospect, you have learned something real about your own targeting rather than generating a longer list.

What you get back

Your network in a queryable table: name, company, title, contact details where they exist, in or out, and why. Plus a weekly shortlist of ten with drafted notes, staged, waiting for you to send them.

Where this goes wrong

Enriching everyone. The single most expensive mistake available here. A network of five thousand contains a lot of students, retirees and people from a previous career. Prefilter first.

Skipping the suppression step. It runs before ranking for a reason. Rank first and you will surface someone you messaged three weeks ago, which is worse than not contacting them at all.

Automating the send. It is the one hard rule in this play. Everything upstream is automation; the message is yours.

Questions

Why does it refuse to give a fit score out of a hundred?

Because a score is a threshold you can move, and everyone eventually moves it until the list is big enough to feel productive. A binary in or out with a written reason forces the definition to do the work. If too much of your network is coming back in, the honest conclusion is that your definition of a prospect is too loose, and that is worth knowing.

Why no LinkedIn automation?

Because it puts your account at risk, and your account is the asset the entire play depends on. Automated connecting and messaging is against the platform's terms and is enforced against. The research and the drafting are automated because they are the slow part. The send stays manual, which also keeps the daily volume in a range that looks like a person.

How many invitations a day is safe?

Around twenty to twenty-five, and the play stages the shortlist with that in mind. The constraint is not really the platform limit, it is that this only works while the messages are personal. Ten well-researched notes will beat a hundred generic ones, and they will not cost you the account.

What about people I am already connected to?

They do not get a connection request, obviously. They get a warm direct message instead, and ideally from whoever in your business actually knows them. The play separates the two routes rather than treating an existing relationship as a cold one.

My export has almost no email addresses in it.

That is normal. LinkedIn only includes an email where the connection has allowed it. The export is useful for who is in your network, not for how to reach them. That is what the enrichment step is for, and it is why the prefilter runs first.