← all playbooks

Score a longlist against the rules your client never wrote down

The same longlist, ranked against the written spec and the unwritten rules, with the evidence quoted beside every score, a reason on every reject, and a column of questions that only a call can answer.

use when
you have a longlist and a spec, but the rules that really decide the shortlist were said on a call and never made it into the document
starts from
A listA job spec

The prompt

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

paste this into Claude
Use hyreflow and score my longlist against the rules my client never wrote
into the spec.

Longlist: <FILE_OR_LIST>. Spec: <PASTE IT, OR GIVE A FILE PATH / URL>.

Ask me for the unwritten rules in plain words before you score anyone. To
start: <RULE_ONE>, <RULE_TWO>, and the commute limit is <COMMUTE_LIMIT>.

How I want it run:
- Read each rule back to me and say how you will test it: from the dated
  work history, from facts about the employer, or not from a profile at all.
- Pull real work history before you score. A job title is not evidence.
- Quote the evidence beside every score. Where a rule cannot be judged from
  a profile, say so and put it in an "ask on the call" column.
- Never score on age, gender, family status, nationality or anything like
  them, whatever the client said.

Run five people first and tell me what the full run costs before you do the
rest. Do not contact anyone.

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

What you need first

  • A longlist with a LinkedIn profile link on every row
  • The job spec, as a file, a link, or pasted text
  • The rules your client said out loud, in your own words
  • A Hyreflow workspace with credits

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 spec and the list

    free

    The written must-haves come out of the spec, and the list is checked for what each row already carries. A row with a profile link and no work history is the normal case.

  2. 2

    Ask for the unwritten rules

    free

    You say them the way the client said them. Each one is read back as a test and sorted: answerable from dated work history, answerable from facts about an employer, or not answerable from a profile. Anything that turns on a protected characteristic is refused here.

  3. 3

    Pull the dated work history

    credits

    Employers, titles and dates for every person on the list, so tenure, progression and the size of company someone has worked in are visible. Rows that already carry a history are skipped, and a provider that returns only a name and a headline counts as a miss.

  4. 4

    Look up the employers a rule turns on

    credits

    A rule about company size or a named group of employers needs facts about the employer. Those are looked up once per company, not once per person, and only for the companies a rule depends on.

  5. 5

    Score with the evidence quoted

    credits

    Every person is judged against the written must-haves and each unwritten rule as met, not met or unknown, with the line of history that supports it. Missing data is recorded as unknown. It is never counted as a fail.

  6. 6

    Fill the ask-on-the-call column

    free

    Clearance, notice period, pay expectations and whether someone would really make the commute do not live on a profile. They are written out per person as the questions to ask, so they are not guessed at.

  7. 7

    Hand over the ranked list and the rejects

    free

    One file ranked by fit, one file of rejects with the rule that ruled each person out. Nothing is sent and nothing is written to your ATS.

What a run costs

Credits are spent per candidate the play actually works, and a lookup that finds nothing usually costs nothing. The two figures are the run where the first provider answers and the run where every lookup walks its full chain.

candidatesif the first provider answersif every lookup walks the chain
25$1.616 credits$3.131 credits
100$6.565 credits$13130 credits
500$33330 credits$63630 credits
1,000$65650 credits$1251250 credits

Free before anything is charged

  • Read the spec and the list
  • Ask for the unwritten rules
  • Fill the ask-on-the-call column
  • Hand over the ranked list and the rejects

What moves the number

  • Coverage on work history. The chain stops at the first provider that answers, and only that provider bills.
  • How many candidates survive the free filters. Everything dropped before the paid steps costs nothing.
  • The scoring and drafting steps run on the metered agent, charged on what they read and write rather than per candidate, so they sit outside this table.
  • Providers you connect with your own key. Those calls bill your account, not your credits.

An estimate, not a quote, priced at the volume credit rate. Your agent sizes the run against your own workspace and tells you what it will cost before it spends anything.

The spec is the half of the brief that got written down

Every recruiter has sent a shortlist that matched the spec line by line and had it bounced straight back. The client was not being difficult. The rules that decided it were said on the briefing call: nobody from the big primes because they are wrong on cost and culture, someone whose career makes sense in a company of two hundred people, a clearance that has to be live on day one, a commute the client will not stretch.

None of that is a search filter. A database matches titles and locations. It cannot tell a person who has spent twelve years inside one large group from a person who has moved through three scale-ups, because both carry the same title today. The difference is in the dated history, and that is why this play pulls it before scoring anyone.

It differs from qualifying a company list against your ideal client, which judges companies. This one judges people, against rules you state in plain language.

What you get back

A ranked file with one row per person: the written must-haves and each unwritten rule marked met, not met or unknown, the sentence of work history behind each mark, an overall score with a one-line reason, and the questions to ask on the call. Beside it, the rejects, each with the rule that ruled them out, so you can argue with any of them.

Variations worth knowing

Tier the rules. Tell the agent which rules are knockouts and which only move the ranking. A knockout removes someone. A preference keeps them on the list, lower down, with the reason shown.

Bring your own notes. If your list came out of your ATS with interview notes attached, say so. A clearance confirmed on a past call is better evidence than anything on a public profile.

Start one step earlier. With no longlist yet, turn a job spec into a ranked shortlist builds one, and these rules can be given to it as extra scoring criteria.

Where this goes wrong

Unknown treated as no. Most profiles say nothing about clearance. If silence counted as a fail, the list would empty itself. Unknown stays unknown and becomes a call question.

A rule that is really a proxy. "Young and hungry" and "no career gaps" are not rules this play will test. Say what the job needs, such as appetite for travel or recent hands-on work, and that can be scored.

Thin profiles. Someone with two lines of history cannot be scored with confidence. They are marked as scored on thin evidence and kept apart from the people scored on real history.

Stale rows. A profile can lag a job move. Check the top of the list by eye before you pick up the phone.

Questions

Why not put these rules into the search in the first place?

Because a search filter knows a title, a location and a current employer. A rule such as a career that fits a company of two hundred people is a judgement on dated history, and a rule such as an active clearance is not on a profile at all. This gate runs after the search, on whatever list you already hold, including one built by hand or exported from your ATS.

What can it not judge from a profile?

Clearance, right to work, notice period, pay, and whether someone would really accept the commute. A profile usually shows a metro area, not a home address, so distance is an estimate at best. These land in the ask-on-the-call column as questions. They are never scored as a pass or a fail.

Will it score on age, gender or family status if my client asks for it?

No. Protected characteristics are never a filter or a scoring input, even when a field in your data exposes them. A rule has to be about skills, experience, the scope of past roles or location. If the client's preference is really about one of those, say that instead and it can be tested.

What does it cost to run?

Asking for the rules, sorting them and writing the call questions cost nothing. The spend is the work history lookup per person, where a miss usually costs nothing, a company lookup per employer where a rule needs one, and the scoring itself, which is metered on usage. Five people run first, and you see what the whole list will cost before it goes ahead.

What if the agent reads one of my rules wrongly?

That is what the read-back is for. Every rule is restated as a test before anything is scored, and every score carries the evidence it was based on. If the big primes turned out to mean a wider group than you intended, you correct the rule and rescore. The work history is already on the rows, so it is not paid for twice.