Most candidate searches turn up plenty of plausible profiles. The harder part is working out who is genuinely worth a closer look for this role, and being able to explain why.

Strawberry is a browser with built-in AI agents. Your companion can work across LinkedIn, portfolios, GitHub, other public work, role documents, and approved ATS context, then show why each person may fit and what still needs checking.

You do not need a perfect scorecard to start. One role and a small, varied candidate set is enough for a useful first pass. Your feedback can sharpen the search before it becomes a list of 100 people nobody trusts.

Start with what makes someone worth finding

Give your companion the best context you have: a job description, scorecard, role notes, examples of strong people, or simply what success should look like in the role. It can turn that into a working definition of fit without making you complete a setup project first.

Keep must-haves separate from useful signals and open questions. If titles, target employers, geography, compensation, or adjacent backgrounds are still unclear, map the talent market before widening the search.

Want to try it?

Ask your Strawberry companion: “Build a calibrated shortlist for this role with evidence-backed fit assessments, then prepare it for internal team review or an approved ATS upload.”

Skill

Source and review candidates for an open role

Start with one role and a small candidate set you can review together.

Skill

Map the talent market for a role

Clarify the talent market when the search itself is still uncertain.

Review a small set before scaling

For a new role or rubric, the first pass works best as a small, varied set: a few strong matches, borderline cases, and people your companion set aside. Reviewing that smaller group makes it much easier to say what feels right, what feels off, and which criterion needs changing before time and credits go into a wider search.

Once the judgment feels right, Strawberry can apply it more widely. If the search stays thin, it can show where the scarcity comes from and what would change by widening a title, location, employer segment, compensation constraint, or evidence bar, instead of quietly padding the list.

Make the fit assessment easy to challenge

A shortlist should tell you more than who matched a keyword. For each person, Strawberry can show:

  • A clear fit assessment: strong potential fit, potential fit with questions, limited fit based on current evidence, or unable to assess.
  • The strongest evidence for that assessment, with source links.
  • Important gaps, conflicting evidence, and confidence.
  • Contactability and ATS state as verified, inferred, unavailable, stale, or not checked.

The assessment helps the team decide who deserves a closer look. It does not prove that someone is qualified, interested, available, eligible, or ready to contact, and it never replaces the hiring decision.

Give the team something useful to review

The shortlist can stay in chat or become a spreadsheet, CSV, structured candidate brief, dashboard inside an approved company tool, or ATS-ready import. A local interactive artifact can also make filtering and comparison easier for your own review.

Agree on the viewers, fields, triage labels, destination, access, and retention before sharing candidate information. Strawberry-published artifacts are publicly viewable, so candidate profiles, contact details, fit assessments, and hiring notes should not be published through that surface.

Skill

Set Up a Shared Team Workflow

Turn the accepted review method into a clear internal workflow for the hiring team.

Move the shortlist into your ATS

Before uploading anything, review the exact candidates, target records, field mapping, create-versus-update behavior, owners, sources, stages or tags, notes, and duplicate handling. An approved shortlist is not automatically approval to write.

For a new mapping, start with a small batch and check the resulting records. The result should show what was created, updated, skipped, failed, or still needs attention before the wider upload continues.

Keep the next step separate

Once the team approves who to contact, your companion can prepare candidate outreach grounded in verified professional context. Reviewing the shortlist does not authorize a message, follow-up, or ATS change.

Skill

Draft candidate outreach

Draft candidate outreach after the shortlist or candidate segment is approved.

When the search works, save the accepted criteria, evidence, fields, and review points as a skill. Share the approved result, the sourcing method, or the wider Recruiting companion only when each would genuinely help the team.

How to map a talent market with AIClarify titles, employers, geography, compensation, and search levers before a large search.
Official Strawberry skill
Copy skill file Download skill file
name
source-candidates
description
Find and calibrate a reviewable candidate shortlist for an accepted role definition or talent market. Use when a user needs more qualified candidates, wants to test a sourcing strategy, or wants evidence-backed candidate research. Do not use for application review, market mapping without a shortlist, or sending outreach.

Source Candidates

Build a defensible shortlist, not a long list. Keep sourcing, team review, ATS changes, outreach, and hiring decisions as separate states.

1. Confirm the search

Use the accepted job description, role profile, scorecard, talent-market map, hiring notes, and prior feedback already available. Confirm only what materially affects the search:

  • outcomes, job-related must-haves, useful adjacent backgrounds, and evidence of capability;
  • location, working pattern, compensation, eligibility, timing, and exclusions; and
  • requested scale, sources, fields, destination, and review point.

If feasibility, titles, target employers, geography, compensation, or adjacent backgrounds remain uncertain, use strawberry/recruiting/map-a-talent-market before scaling. Never use protected or irrelevant personal characteristics as criteria.

2. Calibrate a small, varied set

When the judgment is new or subjective, propose a small set containing strong matches, borderline cases, and people set aside. Explain the planned sources, breadth versus depth, fields, likely scale, and review point. This lets the user correct the search before more time and credits are spent. Reuse a trusted search without unnecessary reconfirmation, but pause before a materially broader, deeper, or more expensive run.

Search approved professional sources where relevant evidence lives, including LinkedIn, portfolios, GitHub, talks, company pages, and approved internal systems. Work sequentially and at a human pace on interactive services; stop on warnings, challenges, rate limits, or unexpected states.

3. Research and assess fit

Require evidence tied to the accepted criteria for every person. Verify identity and current role where material, preserve links, and keep observed facts separate from candidate claims and inference. Label stale, incomplete, or contradictory evidence.

Give each person understandable fit guidance:

  • strong potential fit — current evidence supports most critical criteria and shows no material conflict;
  • potential fit with questions — relevant evidence exists, but important points need checking;
  • limited fit based on current evidence — important criteria lack support or have conflicting evidence; or
  • unable to assess — reliable professional evidence is too incomplete or ambiguous.

Show the rationale, strongest evidence, gaps or counter-evidence, confidence, and sources. These labels help humans choose what to inspect; they are not proof of qualification, comparative rankings, hiring recommendations, or automatic dispositions. If priority tiers would help, define their job-related criteria and keep them reviewable.

4. Check coverage and state

Deduplicate across sources and available ATS, CRM, spreadsheet, referral, or prior-work context. Report:

  • source coverage and blind spots;
  • whether each candidate is new, known, previously contacted, or unresolved;
  • contactability as verified, likely, unavailable, or not checked;
  • ATS state as observed, inferred, stale, unavailable, or not checked; and
  • scarcity signals and the constraints or evidence standards behind them.

Do not imply that a public profile is interested, available, qualified, eligible, or contactable. Never invent people, URLs, roles, contact details, or evidence. Do not weaken the bar silently to reach a requested count.

5. Deliver the shortlist

Provide:

  1. the accepted search definition and caveats;
  2. the shortlist with fit guidance, criterion evidence, sources, confidence, contactability, and ATS state;
  3. a separate review set with the human judgment required;
  4. coverage, duplicates, and unresolved identity or state issues;
  5. a scarcity diagnosis when the requested result could not be met; and
  6. search levers and tradeoffs without changing them automatically.

Use a compact table in chat for a small result. For team review, offer an internal spreadsheet or CSV, structured candidate brief, dashboard in an approved company tool, or ATS-ready import. A local interactive artifact can support filtering and comparison, but candidate data must not be published through a publicly viewable Strawberry artifact.

6. Share internally or prepare the ATS

Creating an artifact, sharing it internally, and uploading records are separate approvals.

Before internal sharing, confirm the reviewers, candidates, fields, triage labels, destination, access, exports, and retention. Use the minimum candidate data needed. If the destination cannot enforce the agreed access, deliver a file for the user to place in an approved company tool. Keep comments and triage state distinct from ATS truth.

Before an ATS upload, show the exact candidates, target records, field mapping, create-versus-update behavior, owners, source attribution, stages or tags, notes, duplicate handling, and blank fields. Require explicit approval. For a new mapping, start with a small batch, verify the resulting records and links, and report created, updated, skipped, failed, and unresolved candidates.

7. Continue deliberately

A human owns advancement, rejection, and hiring judgment. After shortlist approval, use strawberry/recruiting/send-candidate-outreach for draft-first messaging with exact send approval.

When the method proves useful, offer to save its criteria, sources, fit logic, fields, artifact, destination, and review points as a custom skill. Sharing the artifact, the team skill, or the full companion are separate choices. Use strawberry/operations/set-up-shared-team-workflow when the team should reuse the method. A Routine may refresh an internal review file or prepare candidates for review after the workflow is stable; it does not publish candidate data, change ATS state, contact anyone, or make hiring decisions.