Candidate sourcing becomes useful when the shortlist explains why each person fits. A longer list is not automatically a better one, especially when profiles are duplicated, evidence is shallow, or the hiring bar was never made explicit.

Strawberry can search approved professional sources, connect profiles with relevant work, and organize the result around the criteria your team actually uses. Begin with a small sample so your judgment shapes the wider search.

Define the hiring bar before searching

Separate hard requirements such as location, eligibility, compensation, and must-have experience from softer signals such as seniority, technical depth, startup experience, or communication evidence. Existing scorecards, job descriptions, prior searches, and examples of strong or rejected candidates can sharpen the rubric when they are available.

Use examples to clarify the evidence that matters without turning one person's biography into the requirement. Keep protected or irrelevant personal characteristics out of the search.

Calibrate a small sample before scaling

Search a varied first group across the approved sources and review it before expanding. Fewer than ten candidates is often enough to reveal that a criterion is too broad, too narrow, or being interpreted differently than the hiring manager intended.

If nobody in the sample meets the bar, stop and clarify. Scaling a weak search only produces a larger weak list.

Find 30 exceptional candidates

Use this to define the rubric, calibrate the first sample, and build a sourced shortlist.

I want to find 30 exceptional candidates for an open role.

Research the strongest candidates deeply enough

For a larger pool, collect inexpensive signals first and deepen only the strongest matches. Relevant evidence may come from LinkedIn, GitHub, portfolios, personal sites, talks, posts, or other public professional work.

Cross-reference the ATS or tracker before presenting the shortlist. Check identity, duplicates, prior contact, and current pipeline state, and keep real source links beside every material fit claim.

Make the shortlist easy to challenge

Deliver the fields the team needs, an overall assessment, evidence against the important criteria, caveats, confidence, and the number of profiles reviewed and filtered. Include a few candidates set aside and why, especially early in the process, so the team can correct the judgment.

Once the team accepts the rubric, sources, fields, and review method, preserve them as a custom skill. Draft outreach only for an approved shortlist, keep it grounded in verified professional context, and leave messages reviewable until sending is explicitly approved.

Reach out to a candidate shortlist in my voice

Use this after shortlist approval to draft relevant candidate outreach.

Help me contact this candidate shortlist with messages that sound like me and fit each relationship. Start with a representative set I can review before anything is sent.

Make a hiring workflow reusable across my team

Use this when the accepted sourcing process should be shared across the hiring team.

I want to turn an accepted hiring workflow into a clear shared process for my team.
Official Strawberry skill
name
source-candidates
description
Set up and run a candidate-sourcing workflow for an open role or project. Use when a user wants a sourced shortlist based on a role scorecard, examples of strong people, technical evidence, location, availability, or other job-related criteria, especially when creating or refining their own sourcing process.

Source Candidates

Use this Official Skill as starting guidance for a first or redesigned sourcing workflow. If the user already has a tailored sourcing skill, follow that instead.

Context, setup, and planning

Try to understand the user's hiring setup before researching widely. Approved job descriptions, scorecards, ATS history, prior searches, hiring documentation, and examples of strong or rejected candidates can answer questions faster when they are available. Relevant context may include:

  • the role, responsibilities, and evidence of success;
  • hard criteria such as location, compensation band, must-have skills, and eligibility;
  • softer signals such as seniority, hypergrowth experience, past founder roles, technical background, or communication evidence;
  • sourcing preferences such as LinkedIn, LinkedIn Recruiter, GitHub, Hacker News, niche communities, referrals, or an approved sourcing tool;
  • the user's existing sourcing and evaluation approach, examples worth preserving, and where candidates are stored;
  • required candidate fields, preferred outreach approach, destination, and review point;
  • whether the user wants a broad surface scan or deeper research on fewer candidates. Explain that a common pattern is a wide, lower-cost pass followed by deeper research on the strongest matches.

Criteria can be vague on a first run and sharpen through review. Examples should calibrate the search without turning one person's biography into the requirement. Never use protected or irrelevant personal characteristics as criteria.

When the rubric is not already clear, offer to calibrate from approved examples, a small and varied set of real candidates Strawberry finds, or both. A surface-level sample, often fewer than ten candidates, can reveal disagreements before a larger run. If nobody in the sample meets the rubric, stop and ask rather than scaling a weak search. Never fabricate candidate profiles for calibration.

Present a concise plan covering the rubric, sources, search depth, calibration method, intended deliverable, destination, and review point. Let the user confirm or adjust it before substantial research. Before scaling beyond a calibration sample, restate the scope and expected depth or cost.

Execution

  1. Turn the accepted requirements into a concise, job-related rubric. Keep hard requirements distinct from softer signals so the user can correct either.
  2. Search the user's preferred public professional sources. Separate overlapping searches so the same ground is not covered repeatedly.
  3. For a large pool, extract once using inexpensive signals, narrow the pool, then enrich the strongest subset. For a requested deep dive, go beyond the profile into relevant portfolios, GitHub work, personal sites, talks, posts, or other public professional evidence.
  4. Cross-reference the ATS, candidate sheet, or other approved destination and prior sourcing context to avoid people already in the pipeline or already contacted.
  5. Use real source links and specific evidence for every inclusion, exclusion, and fit claim. Flag low confidence rather than guessing identity, background, experience, or contact details. Do not infer protected or sensitive characteristics.
  6. Follow the agreed calibration method. Review the initial set with the user, incorporate their corrections, then apply the sharper rubric to the wider pool.
  7. Before presenting, check for duplicates, identity mismatches, and stale pipeline state. Rank the shortlist with an overall rating and evidence against the important criteria. Report how many profiles were reviewed and filtered, and, especially early on, include a few candidates set aside and why so the user can calibrate the judgment.

Suggested outcome

Produce the agreed, reviewable candidate-sourcing result in the requested destination and format. It may be a shortlist, ATS update, research brief, spreadsheet, table, or another artifact. Include real source links, requested fields, an overall rating, per-criterion fit evidence, caveats, confidence, and reviewed and filtered counts. Include useful outreach context or profile and work images only when requested or helpful to the agreed deliverable. Never invent experience or contact details.

Suggested next steps

  • If the user wants to contact the approved shortlist, use strawberry/sales/send-personalized-outreach.
  • If a team should reuse the accepted scorecard, fields, and review process, use strawberry/operations/set-up-shared-team-workflow.
  • After review, offer to save the accepted rubric, sources, calibration approach, fields, destination, and review process as a user-owned sourcing skill.
  • After a few successful runs, when the user's preferences are clear, offer a recurring weekly or bi-weekly sourcing pass if the role has enough volume to justify it.