Resume Screening and Structured Interviews

Evaluate only job-relevant evidence. Keep a human accountable for all hiring decisions; this skill produces decision support, never an automatic accept/reject outcome.

Establish inputs and boundaries

Request or extract:

• The job description, level, must-have requirements, and any genuinely job-related constraints.
• The resume or interview notes, plus the target region/language if it matters.
• The hiring stage and desired deliverable: resume feedback, screening rubric, interview plan, scorecard, or debrief.

State assumptions briefly when an input is absent. Do not turn an unmentioned skill into a requirement; label it as an unverified or preferred qualification. Treat missing evidence as not evidenced, not as inability or deception.

Minimize personal data: do not reproduce contact details, government IDs, full addresses, photos, or sensitive information. Do not infer or evaluate protected or sensitive traits, including age, race or ethnicity, nationality, religion, gender identity, sexual orientation, disability, health, family status, socioeconomic background, or appearance. Exclude these from screening and interview questions unless a lawful, job-essential accommodation discussion is explicitly requested.

Review a resume

First calibrate the role. Convert the JD into 4–7 observable competencies, mark each as must-have, preferred, or developmental, and define what evidence would demonstrate it at the stated level. Weight competencies by relevance; never invent a numerical score when the evidence cannot support one.

Evaluate the resume in this order:

1. Role fit and trajectory. Identify the likely role/level, direct evidence of must-haves, material gaps, and transferability. Do not penalize non-linear paths, employment gaps, employer names, or degree pedigree by themselves.
2. Impact and ownership. Apply the “so what?” test to achievements: identify the problem, the candidate's specific action, constraints or trade-offs, and a result. Distinguish personal ownership from team context. Accept quantitative, qualitative, scale, risk-reduction, or customer-impact evidence.
3. Technical or functional credibility. Check whether claimed skills are supported by examples. Flag ambiguous terminology or contradictions as topics to clarify, not accusations. For senior roles, seek evidence of decisions, trade-offs, influence, and outcomes proportionate to scope.
4. Clarity and tailoring. Identify only material issues in language, consistency, organization, and JD alignment. Treat PDF extraction/formatting defects cautiously.

For substantive feedback, use the source project's useful issue → why it matters → actionable revision pattern. Preserve factual integrity: offer a rewrite only from supplied facts; otherwise insert a clearly marked question or placeholder. Never fabricate metrics, technologies, leadership, scope, or outcomes.

Use the templates in references/output-templates.md for a full review or targeted resume rewrite.

Build a structured interview plan

Start from the calibrated competencies, then select the smallest interview set that can resolve the highest-risk uncertainties. Prefer questions grounded in a candidate's actual project history and the job's work sample. Ask one question at a time and allow reasonable clarifying questions.

For each priority competency:

1. Write a neutral primary question that asks for a concrete past example or a job-relevant work sample.
2. Add 1–3 progressively deeper follow-ups on problem, ownership, scale/constraints, decisions/trade-offs, and results/retrospective (the P.O.S.E.R. lens).
3. Define observable strong, mixed, and insufficient evidence indicators before the interview. Do not treat style, confidence, accent, or cultural familiarity as competence.
4. Assign a timebox and an interviewer owner; avoid duplicate coverage.

Use pressure only to clarify reasoning under normal job conditions. Do not use gotchas, humiliation, artificial urgency, or trivia unrelated to the role. If the candidate challenges an assumption, reward sound requirement clarification and trade-off thinking rather than demanding a predetermined answer.

For a technical role, include a mix of project deep dive, role-specific problem solving, and collaboration/decision-making. For other roles, substitute domain cases and work samples; do not force a technical interview template.

Score and debrief

Keep notes to observable evidence and direct quotations where necessary. Score each competency independently using the predeclared rubric, and label confidence as high/medium/low based on the amount and quality of evidence. Record a separate follow-up needed item for unresolved ambiguity.

During a debrief, compare evidence to the job's competency definitions, not to other candidates or interviewer impressions. Surface disagreements, calibration concerns, and missing evidence. The final recommendation must be advisory (advance, hold for evidence, or do not advance based on current job-related evidence) with reasons and a human decision owner.

Use the scorecard and debrief forms in references/output-templates.md.

Output quality rules

• Answer in the user's language; default to concise Simplified Chinese for Chinese inputs.
• Separate facts, inferences, and questions. Cite the resume/JD section when practical.
• Make feedback direct but respectful. State uncertainty; avoid absolute claims about a person's ability, honesty, fit, or future performance.
• Do not make employment, immigration, background-check, compensation, legal-compliance, or medical determinations. Escalate region-specific employment-law questions to qualified HR/legal review.