31 Claude Skills for Recruiting

What these skills are

One search from intake to offer, as installable Claude skills. The skills follow one search from start to finish. You brief the role with the hiring manager, keep candidates informed at every stage, attract the right applicants, source and reach out, screen and interview on the same criteria for everyone, then decide and close the offer. Agency recruiters get a desk of their own, and the last three skills run it: the numbers, the rules for AI, and the interviewers.

The red line: Claude writes the search, the questions and the message, never the verdict. It does not screen, rank or score a candidate, and a person reads every application and makes every hiring decision.

Each skill is one named method or one artifact you walk away with: an intake summary, a job scorecard, Boolean strings, an outreach sequence, a blank interview scorecard, a debrief record, an offer letter, a client search update. Each works from what you paste — the notes from the intake call, the profiles the manager liked, the reasons a client gave — with no ATS connector to install, and each one ends with the decision a named person makes, by a date. Pay transparency, work authorisation, automated decision rules and agency terms are sent to a qualified adviser rather than answered.

Mechanically, each skill is one folder with a SKILL.md file. The 31 are grouped into eight parts of a search, and each one gives full value on its own: run one, run a part, or work through the lot in the order the search moves.

Download all 31 skills. One zip: ready-to-install skill zips, readable SKILL.md files, and the evidence notes behind every method. Free, no signup. The same 31 skills are open on GitHub: polar-bear-org/claude-skills.

The 31 skills you get

The set follows the search, in eight parts: briefing the role, keeping candidates in the loop, attracting, sourcing and reaching out, screening and interviewing, deciding and closing, the agency desk, and running the desk.

1 · Brief the role

1. Intake Meeting

Use when: The manager cannot describe the job and your search drifts

Output: A question plan for the hiring manager, a calibration on three to five anonymised sample profiles, a must-have versus trainable list, and a one-page intake summary to sign off

2. Job Scorecard

Use when: Two jobs hide in one title and "senior" means five things

Output: First-year outcomes, four to six job-related criteria with what evidence of each looks like, the must-have versus trainable split, and the stage where each criterion is assessed

3. Salary Range Brief

Use when: You fear "we didn't budget for this" after four rounds

Output: The proposed range from the evidence you supply, where the range must appear, what to say to candidates above or below it, and pay transparency questions for an adviser

4. Hiring Process Plan

Use when: There is always one more interview and the pipeline goes cold

Output: A stage map with the criterion assessed at each step, a maximum number of rounds, feedback and decision deadlines, the decision meeting date and the candidate touchpoints

2 · Keep candidates in the loop

5. Candidate Communication Plan

Use when: Candidates hear nothing for weeks and the good ones go silent

Output: A promise per stage naming who writes and by when, the "no update yet" message, a close-the-loop rule, and keep-warm notes from offer to first day

6. Interview Scheduling Messages

Use when: A screen was cancelled twelve minutes before it started

Output: An invitation that says what to expect and offers adjustments, then confirmation, reminder and reschedule, and apologies for a late cancellation or a missed slot

7. Rejection Email

Use when: You have decided no and need to say it kindly and fast

Output: Messages per stage with feedback lines drawn only from the documented criteria, a short phone script for finalists, and a calm reply to a hostile response

8. Candidate Experience Survey

Use when: Strong candidates decline because of how the process felt

Output: Five to seven questions per stage, when to send it, how to read the results in aggregate, and a fix list for the process owner

3 · Attract

9. Job Description

Use when: You are about to repost last year's description

Output: An outcomes-first description built from the scorecard, must-haves only, a pay range line, an adjustments line, and a wording check that flags gendered, age-coded and exclusionary words

10. Job Ad

Use when: Hundreds apply who ignored the rules, and the right people scroll past

Output: A two-line hook, the real job including its hard parts, pay, location and work authorisation rules up front, how to apply, and what happens next by when

11. Employee Referral Program

Use when: Inbound is full of fakes and you want warm introductions

Output: The ask to staff built from the scorecard, a guide to what makes a good referral, bonus rules as placeholders, the same assessment for referred people, and a thank-you and update loop

4 · Source and reach out

12. Boolean Search Strings

Use when: Your strings return thousands or nobody

Output: Title and skill synonym sets, strings for LinkedIn, Google X-ray and job boards in a broad and a narrow version, and a test-and-tighten log

13. Talent Mapping

Use when: The manager wants "our exact industry" and you need evidence

Output: The target companies and teams where the profile works, adjacent titles, feeder fields and a pool size estimate from public information, with no files on individuals

14. Sourcing Strategy

Use when: Inbound is unusable and replies have collapsed

Output: A channel mix across outbound, referrals, communities, past applicants, boards and events, free routes when you have no paid seat, weekly targets per channel, and a list of what to stop

15. Recruiting Email Templates

Use when: Everyone can spot a message pasted from a chatbot

Output: A first message and two follow-ups built only from facts you supply, with subject lines, a short InMail version, and a what-not-to-say list

5 · Screen and interview

16. Screening Questions

Use when: Hundreds of applications a week and a person still reads each one

Output: Evidence questions for the application form in place of knockouts, a phone screen guide asked the same way to everyone, and an identity and work-authorisation step done by a person

17. STAR Interview Questions

Use when: Candidates answer "tell me about a time" with what they would do

Output: Behavioural questions per scorecard criterion, STAR probes for vague or "I would" answers, one situational question per criterion, and a never-ask list

18. Structured Interview

Use when: Every interviewer asks their favourite questions

Output: The panel split of criteria, the running order with timings and fixed probes, the opening and closing scripts, note-taking rules, and what the candidate is told in advance

19. Work Sample Test

Use when: CVs no longer tell you who can do the work

Output: A short task that mirrors the job, the time limit and payment decision, the candidate instructions, and an evidence guide interviewers use to mark the work

20. Interview Scorecard

Use when: You need evidence written down before anyone says "I liked her"

Output: One blank form per interviewer, a section per criterion with anchored descriptions of weak to strong evidence, a notes box before each level box, and no total score

6 · Decide and close

21. Interview Debrief

Use when: The loudest voice decides, or the manager asks to "see a few more"

Output: An agenda that collects written scorecards before anyone speaks, a walk through the evidence criterion by criterion, and a record of the hiring manager's decision and reasons

22. Reference Check Questions

Use when: You want references that tell you something

Output: A consent step, questions tied to the scorecard criteria and the debrief's open questions, probes past polite answers, a never-ask list and a note template

23. Offer Letter

Use when: The offer is approved and must go out today

Output: A letter drafted from terms a named person approved, the conditions and start date, an offer call script, and a clause list for adviser review

24. Counter Offer Plan

Use when: Declines and counter-offers keep taking your accepted candidates

Output: Questions to ask before the offer, resignation conversation prep, honest answers when the current employer counters, and a keep-warm plan from notice period to day one

7 · Agency desk

25. Recruiter Business Development

Use when: Mass outreach to decision makers is dead

Output: A target account list from hiring signals you gather, a one-page approach per account, first-call questions, and a four-touch outreach plan

26. Agency Terms of Business

Use when: Clients ignore fees or hire someone you introduced

Output: Each term in plain language — fee basis, retained versus contingency, guarantee or rebate, introduction period, payment terms — a one-page client summary, and a clause checklist for an adviser

27. Candidate Submittal

Use when: You want the client to read evidence, not adjectives

Output: A consent-to-represent record, the facts against each scorecard criterion in the candidate's own words, and motivation, pay expectation, notice and availability, with no fit score

28. Client Search Update

Use when: The client goes quiet, chases daily, or rejects everyone

Output: Activity by stage, what the market is saying about why candidates say no, a proposed brief change when the evidence supports it, and decisions requested by a date

8 · Run the desk

29. Recruiting Metrics

Use when: You are judged on roles that were dead on arrival

Output: Time to fill and time to hire, source of hire, offer acceptance, pass-through by stage, and requisition ageing with the reason each role is stuck, at requisition and channel level only

30. AI Recruiting Policy

Use when: A vendor promises AI screening and you need a line to defend

Output: Allowed and banned uses, the candidate notice text, the human review steps, and the questions to put to vendors and a qualified adviser

31. Interviewer Training

Use when: Interviewers improvise and the debrief has no evidence

Output: A 45-minute session for hiring managers and panels on the structured interview, note-taking, independent scorecards and questions never to ask, with a one-page cheat sheet

Where to start, and what the pack rests on

There is no compulsory sequence. You bring one live search — the notes from a twenty-minute call with the hiring manager, the three profiles she liked, the reasons a client gave for turning everyone down — and open the skill that matches where the search is stuck. The order in the catalogue is the order a search moves in: briefing the role first, because a brief nobody agreed on is what makes every later step drift, then the promise to candidates, which is the thing that breaks first when a search runs long.

Most of them chain. recruit-intake-meeting produces the agreed brief that recruit-job-scorecard turns into criteria, and the scorecard is then the input to nearly everything downstream: recruit-job-description and recruit-job-ad write from it, recruit-boolean-search builds its synonym sets from it, and recruit-screening-questions, recruit-star-interview-questions, recruit-structured-interview and recruit-interview-scorecard all assess the same criteria, stage by stage. Closing runs in order too: recruit-interview-debrief collects the scorecards and records the decision, recruit-reference-check picks up the questions the debrief left open, recruit-offer-letter goes out on approved terms, and recruit-counter-offer holds the accepted candidate to day one. On the agency desk, recruit-business-development wins the client, recruit-terms-of-business agrees the terms, recruit-candidate-submittal presents the evidence and recruit-client-search-update reports on the search.

The pack is research-informed, not validated. A scan of what recruiters and candidates say they struggle with set the order of the skills and the words in each "use when": 198 short pain statements collected from public posts, forum threads and question-shaped searches on 26 and 27 September 2026, grouped into twelve struggles. Reddit leans to venting and US posters, LinkedIn leans to leaders and vendors, and relevance search is not a random sample, so the counts show emphasis in what was pulled, not how common a problem is across the profession. The methods themselves come from public guidance, institution pages and peer-reviewed selection research — US OPM on job analysis, structured interviews, work samples and reference checking, Levashina and colleagues (2014) on what interview structure is made of, Sackett and colleagues (2022) on the validity of selection methods, Flanagan (1954) on critical incidents, Janz (1982) and Latham and colleagues (1980) on behavioural and situational questions, Smith and Kendall (1963) with Google re:Work on anchored rating scales, Gaucher and colleagues (2011) on gendered wording and Premack and Wanous (1985) on realistic job previews, CIPD and SHRM on selection methods and referrals, the US EEOC and the Uniform Guidelines, LinkedIn Recruiter and Google Search help on Boolean syntax, Bain's Net Promoter System, the EU AI Act and the pay transparency directive, New York City's automated employment decision tool rules and the UK ICO's recruitment guidance, and the UK agency conduct regulations with the American Staffing Association's code. The one shared file in the pack, resources/evidence-and-sources.md, lists them source by source with the finding used and the limit on it. No skill here has been tested in a trial.

Setup guide

1. Download the pack. One zip: an install folder with 31 ready-to-upload skill zips, a skills folder with the same 31 skills as readable SKILL.md files, and the evidence notes — resources/evidence-and-sources.md — which name the source behind every method and where it stops.
2. Install your skills. In Claude Code, add the marketplace and install ai-for-recruiting-pack, and all 31 load at once. In Claude, turn on code execution in Settings, Capabilities, then go to Customize, Skills and upload one zip per skill from the install folder. Prefer working from files? Add the SKILL.md files to your Project knowledge instead; it works, just less cleanly.
3. Bring one live search. No connector, no ATS integration, no admin: every skill works from what you paste — your notes from the intake call, the three profiles the manager liked, the reasons a client gave for turning everyone down. Start with the search that is stuck and write "run recruit-intake-meeting". Where you have little, each skill starts from the minimum and marks what is still to find.

Where to start

The quality bar

Every skill in the pack holds the same standard, the one we hold when we hire our own people:

• Every skill starts from the job: outcomes and criteria agreed with the hiring manager, not a description copied from last year
• Every candidate meets the same questions, the same stages and the same messages
• Every output has a fixed shape, a finish line, and a named person who decides next and by when
• Messages use only the facts you supply: nothing is scraped, enriched, invented or guessed about a person
• Legal and regulatory points — pay transparency, work authorisation, automated decision rules, agency terms — are flagged for a qualified adviser rather than answered
• No skill screens, ranks or scores a candidate: scorecards ship blank and carry no total, metrics stop at requisition, stage and channel level, and talent maps stay at company and team level
• The red line: Claude writes the search, the questions and the message, never the verdict — a person reads every application and makes every hiring decision

Who made this

Polar Bear is a people ops consultancy for human-size teams (20 to 200 people). Built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. This pack is the free, self-serve version of how we work.

The pack carries one recruiter's search. When you want your whole team working this way, AI carrying the overhead so people do the part only people can do, across hiring, management and everyday operations, that's what we build with clients.

Meet Pauline. Hiring for a role nobody can describe, or losing candidates at the offer? Bring the stalled search and the manager who wants more CVs. Book a 30-minute call · Pauline on LinkedIn

Install in Claude Code

Two lines, and every skill in the pack loads at once.

/plugin marketplace add polar-bear-org/claude-skills
/plugin install ai-for-recruiting-pack@polar-bear-skills

Using Claude on the web instead? Download the zip and upload each skill from its install folder under Customize → Skills.