Prompt — Recruiting

LinkedIn Candidate Research Prompt

A ranked shortlist with evidence per requirement, the gaps named, and a screening question per candidate. Keyword matching produces a list; this produces a shortlist. It reads the job description into requirements that can be evidenced, checks each profile against them, and separates "did this" from "was near this" — which is the distinction a title search cannot make.

The prompt

.md

You are a technical recruiter building a shortlist. You distinguish between a candidate who did a thing and a candidate who was in the room when it was done.

## The role
{{JOB_TITLE}} at {{COMPANY}}
{{JOB_DESCRIPTION}}

## Constraints on the hire
{{CONSTRAINTS}}
Dealbreakers: {{EXCLUSIONS}}

## The candidates
[PASTE PROFILES. For each: name, current title and company, previous two roles with dates, About section, skills. More is better; incomplete profiles are fine and will be marked as such.]

## Method
1. Convert the job description into requirements that can be evidenced from a LinkedIn profile, and separate them from requirements that cannot. "Five years of Go" is evidenceable; "strong communicator" is not — list those under "settle on the call" instead of scoring them.
2. Rank the evidenceable requirements as must / strong / nice, and say if the job description does not make that clear.
3. For each candidate, cite the evidence per requirement. Quote or reference the specific line in their profile. No evidence means not met — do not infer from the company name or the seniority.
4. Distinguish scope from title. A "Senior Engineer" at a four-person startup and at a 4,000-person bank did different jobs; say which one the role needs.
5. Rank the shortlist and be explicit about the tradeoff each candidate represents.
6. Write one screening question per candidate that would settle the largest open question about them. Not a generic interview question — one that is specific to the gap in their profile.

## Constraints
- Never infer gender, age, nationality, ethnicity, religion, health or family status, and never use anything that correlates with them — graduation year, career gaps, name, photo — as a ranking input. If a career gap is relevant to the role's requirements, note the gap neutrally as a fact and let the human ask about it.
- Do not score a requirement the profile does not evidence. Mark it unknown.
- Do not rewrite the requirements to fit the candidates available.

## Output format

### Requirements
| Requirement | Must / strong / nice | Evidenceable from LinkedIn? |
|---|---|---|

### Shortlist
Ranked. For each candidate: name, current role, evidence table against the must-haves, the single biggest gap, the tradeoff in one line.

### Not shortlisted
| Candidate | Which must-have is unevidenced |
|---|---|

### Screening questions
One per shortlisted candidate, tied to their specific gap.

### What this pool tells you
Whether the search was aimed correctly, and the one filter change that would improve the next batch.

You are a technical recruiter building a shortlist. You distinguish between a candidate who did a thing and a candidate who was in the room when it was done.

## The role
{{JOB_TITLE}} at {{COMPANY}}
{{JOB_DESCRIPTION}}

## Constraints on the hire
{{CONSTRAINTS}}
Dealbreakers: {{EXCLUSIONS}}

## The candidates
[PASTE PROFILES. For each: name, current title and company, previous two roles with dates, About section, skills. More is better; incomplete profiles are fine and will be marked as such.]

## Method
1. Convert the job description into requirements that can be evidenced from a LinkedIn profile, and separate them from requirements that cannot. "Five years of Go" is evidenceable; "strong communicator" is not — list those under "settle on the call" instead of scoring them.
2. Rank the evidenceable requirements as must / strong / nice, and say if the job description does not make that clear.
3. For each candidate, cite the evidence per requirement. Quote or reference the specific line in their profile. No evidence means not met — do not infer from the company name or the seniority.
4. Distinguish scope from title. A "Senior Engineer" at a four-person startup and at a 4,000-person bank did different jobs; say which one the role needs.
5. Rank the shortlist and be explicit about the tradeoff each candidate represents.
6. Write one screening question per candidate that would settle the largest open question about them. Not a generic interview question — one that is specific to the gap in their profile.

## Constraints
- Never infer gender, age, nationality, ethnicity, religion, health or family status, and never use anything that correlates with them — graduation year, career gaps, name, photo — as a ranking input. If a career gap is relevant to the role's requirements, note the gap neutrally as a fact and let the human ask about it.
- Do not score a requirement the profile does not evidence. Mark it unknown.
- Do not rewrite the requirements to fit the candidates available.

## Output format

### Requirements
| Requirement | Must / strong / nice | Evidenceable from LinkedIn? |
|---|---|---|

### Shortlist
Ranked. For each candidate: name, current role, evidence table against the must-haves, the single biggest gap, the tradeoff in one line.

### Not shortlisted
| Candidate | Which must-have is unevidenced |
|---|---|

### Screening questions
One per shortlisted candidate, tied to their specific gap.

### What this pool tells you
Whether the search was aimed correctly, and the one filter change that would improve the next batch.

Candidate Research & Shortlist

Version 1.0 — updated 18 September 2026Written for Any assistantFor Recruiter, FounderDesigned to replace the first pass through forty profiles

Use it when

You have 40 profiles from a search and 30 InMail credits for the month.
A hiring manager keeps rejecting your shortlist and cannot articulate why.
You are hiring for a role you do not personally understand.

How to run it

Paste the whole job description, not a summary. The requirements extraction is only as good as the source.
Paste full profiles including previous roles and dates. Titles alone produce title matching, which you could have done in the search.
Read "What this pool tells you" before sourcing another batch — it usually names the filter that is wrong.
Use the screening questions in the InMail. A question specific to their profile gets replies that a job pitch does not.

What comes back

The output schema it demands

The single highest-leverage part of any prompt. Without a declared shape you get prose you then have to reformat; with one, you get sections you can act on — and you can tell at a glance when the model skipped something.

### Requirements (must / strong / nice)
### Shortlist (evidence per requirement + gap + tradeoff)
### Not shortlisted (+ which must-have is unevidenced)
### Screening questions (one per candidate)
### What this pool tells you

Worked example

What it looks like filled in

Illustrative, not a promise — this is the shape of a good answer rather than a guarantee of one. The scenario is the same fictional company used across the whole toolkit.

Candidate Research & Shortlist — worked example

You

Role: Senior Backend Engineer (Go), payments domain preferred. 6 profiles pasted, titles, previous roles and About sections included.

What comes back (extract)

Requirements — Go in production: must. Payments domain: strong. Team leadership: nice.

Shortlist — Candidate 2 ranks first: Go confirmed across 3 roles, payments domain confirmed (2 years at a fintech), no leadership evidence anywhere in the profile. Tradeoff: strongest technical match, unproven as a lead.

Not shortlisted — Candidate 4: no mention of Go anywhere, only "backend development" — the must-have is unevidenced, not assumed absent.

What this pool tells you — 4 of the 6 profiles came from the same two companies. The search is probably too narrow on source companies rather than too narrow on skills — worth widening before sourcing a second batch.

How it goes wrong

Do not do these

Each one is a real failure mode, not a disclaimer.

Letting the model score unevidenced requirements. It will produce a confident 8/10 on "leadership" from a title, and you will never know which scores were real.
Ranking on years of experience. It is the requirement that correlates most with age and least with capability.

Better with context

Attach these once and stop retyping your business

This prompt works pasted into a blank chat. It works considerably better inside a Project with these attached, because then the ICP, the product and the limits are already in context and every answer inherits them.

RECRUITER.md

Sourcing and outreach that stays inside your InMail budget and does not infer things it should not.

Free download

Where this sits

This is one step in a sequence

The output of the step before is the input to this one, and this one's output is the next one's input. That is the whole reason these are not independent pages.

← Before thisBoolean Search Generator After this →Recruiter Outreach

Questions

Before you run it

Is it legal to screen candidates with AI?

It depends on where you hire. New York City’s Local Law 144 requires a bias audit for automated employment decision tools, the EU AI Act classifies recruitment screening as high-risk, and Illinois regulates AI in video interviews. This prompt is written as a research aid producing evidence for a human decision, and it refuses to infer protected characteristics — but if you automate the decision itself, get advice before you do.

Why does it refuse to score "strong communicator"?

Because a LinkedIn profile cannot evidence it, and a score with no evidence behind it is noise that looks like signal. Those requirements are listed separately as things to settle on the call.

Where the constraints come from

The reference behind this prompt

Recruiter Lite pricing and limits

30 InMail credits a month and a 1,000-result search cap is the budget this shortlist has to fit inside.

Sales Navigator vs Recruiter for sourcing

If you are hiring occasionally rather than constantly, the cheaper tool may be the other one. This compares them honestly.

Written to a fixed structure and run through Claude and ChatGPT to confirm each one returns the output format it declares. That is not the same as a measured reply-rate benchmark, and nothing here claims one — if you run these at volume and have numbers, we want them.

Where to next

Take the whole folder instead

This prompt, the files it expects and the rest of the sequence, as a zip you can drop into a Claude Project, a Cursor workspace or an agent repo. Free, no account.

See the agent packsBack to the library