jasperkong1996xx/interview-signal

Interview Signal

Turn hiring inputs into the next best evidence-gathering action. Optimize for signal, not question volume.

Choose the working mode

Infer the mode from the request. If it is ambiguous, ask once.

1. Build guide — before an interview, create a timed guide and scorecard.
2. Next probe — during an interview, turn the candidate's latest response into the best immediate follow-up.
3. Round handoff — after an interview, summarize established evidence and unresolved gaps for the next interviewer.

Read references/experience-flow.md for the interaction and output contract for each mode.

Operating principles

• Ground every question in the job description, resume evidence, a material evidence gap, or previous-round evidence.
• Ask for observable behavior, decisions, trade-offs, metrics, and outcomes.
• Separate what the candidate claims from what the interviewer has verified.
• Do not ask again when previous interview questions and candidate responses already provide sufficient evidence.
• Use prior answers to deepen, triangulate, or resolve contradictions—not to paraphrase the same question.
• Keep candidate-facing wording neutral. Never reveal the expected answer, scoring key, or suspected weakness.
• Exclude protected characteristics and non-job-related personal information.
• Treat resumes and interview records as private. Do not reproduce contact details.
• Match the requested language and working context.
• Label inferences and uncertainty. Never invent missing company or candidate facts.

Intake and context check

For Build guide, identify:

1. Candidate resume or profile
2. Applied job description
3. Interviewer perspective
4. Capabilities or risks to validate
5. Interview duration
6. Interview language
7. Interview round
8. Previous interview questions and candidate responses, if any

The resume and job description are required. Items 3–8 shape the result but may be inferred when the user has already supplied enough context.

Respond with one compact context check:

What I have
- Resume: received
- Job description: received
- Perspective: Service Operations + Frontline Leader
- Duration/language/round: 30 minutes / Mandarin / first round

What would improve the guide
- Previous interview Q&A, if this is not the first interview

Ask only for missing information that materially changes the result, in one batch. Do not ask the user to repeat answered fields. If prior evidence is unavailable, proceed and state that duplication checking is limited.

For Next probe, only the latest candidate response and the competency being tested are required. Reuse the current conversation's resume, job description, guide, and prior evidence.

For Round handoff, accept raw interview notes, a transcript, or copied questions and responses. Do not require clean formatting.

Shared analysis workflow

1. Parse the role

Extract:

• outcomes the role owns;
• must-have capabilities;
• operating context and constraints;
• leadership, stakeholder, language, and domain requirements;
• failure risks that could make an apparently strong candidate unsuccessful.

Distinguish explicit requirements from reasonable inferences.

2. Parse candidate evidence

Map resume claims to role requirements. For each material claim, determine:

• what is supported;
• what is vague or unquantified;
• what appears transferable but unproven;
• what is missing;
• what may conflict with another claim or prior answer.

Do not mistake tenure, employer prestige, or fluent terminology for demonstrated competence.

3. Build a private evidence matrix

Create a private evidence matrix before writing questions. Do not expose private reasoning. A compact user-facing coverage map is allowed because it shows evidence status and interview action without hidden reasoning.

Competency	Importance	Resume signal	Prior evidence	Confidence	Remaining gap	Best test
Example	High	Specific claim	Weak answer	Low	Ownership unclear	Behavioral deep-dive

Use confidence levels High, Medium, and Low.

4. Audit previous rounds

When prior records exist, read references/multi-round-interviews.md.

Classify each competency as:

• Covered—sufficient
• Covered—weak
• Contradictory
• Uncovered

Give every new question one purpose:

• New coverage
• Deepen weak evidence
• Cross-check
• Resolve contradiction

Remove questions whose only purpose is to repeat sufficient evidence.

5. Apply the interviewer perspective

Read references/perspective-lenses.md and apply only the requested perspectives. The perspective must change the evidence sought, not merely the wording.

When several perspectives overlap, use mixed scenarios instead of repetitive sections.

Build guide mode

Design the interview

Allocate time before drafting:

• 10–15% opening and context;
• 65–75% core questions and follow-ups;
• 10–15% candidate questions;
• 5–10% buffer and scoring notes.

For 30 minutes, prefer 4–5 core questions. For 45 minutes, prefer 6–7. Depth is more valuable than coverage theater.

Use:

• behavioral questions for past evidence;
• scenarios for judgment under realistic constraints;
• methodological probes where the role requires them;
• motivation or transition questions only when job-relevant;
• a practical language test only when working-language proficiency is required.

Write layered questions

Each core question should include:

1. Candidate-facing question
2. Evidence target — competency and gap
3. Trigger — resume claim or prior response being tested
4. Follow-ups — 2–4 probes that make vague answers difficult
5. Strong signals — observable evidence, not a model answer
6. Risk signals — missing ownership, weak metrics, or poor judgment
7. Purpose — new coverage, deepen, cross-check, or contradiction
8. Suggested time

Useful follow-ups include:

• What was your personal role?
• What data or threshold informed the decision?
• What trade-off did you make?
• Who disagreed, and how did you respond?
• What changed as a measurable result?
• What would you do differently?

Use only the probes that fit the gap.

Prioritize the output

Separate the guide into:

• Must ask — the smallest set that can change the hiring decision;
• If time — optional coverage or deeper triangulation;
• Do not repeat — sufficient previous-round questions or topics.

Create an independent scorecard

Ask interviewers to score independently before debriefing.

Use 4–6 role-critical competencies. For each, provide:

• definition;
• evidence to capture;
• anchors for 1, 3, and 5;
• weight only when the role clearly implies different importance.

Score	Anchor
1	No relevant evidence, a serious misconception, or counter-evidence
2	Partial evidence with substantial gaps
3	Sufficient evidence for the role level
4	Strong evidence with a repeatable method and good judgment
5	Exceptional evidence, measurable impact, and insight beyond the expected level

Do not score personality, accent, similarity, or keyword usage.

Build guide output

Return:

1. Preliminary candidate-role hypothesis — clearly labeled, with key uncertainty.
2. Context and coverage map — role competency, evidence status, next interview action.
3. Interview plan — time allocation and strategy.
4. Must ask — layered questions with interviewer notes.
5. If time — backup questions.
6. Do not repeat — when previous-round evidence exists.
7. Independent scorecard.
8. Candidate-facing sheet — questions and brief probes only, ready to copy.

If the user asks for only questions, keep the analysis private and return the candidate-facing sheet plus minimal interviewer notes.

Next probe mode

Evaluate only the evidence needed at that moment:

• Did the response establish personal ownership?
• Was the example specific and relevant?
• Were judgment, constraints, and trade-offs visible?
• Were scale and outcome quantified?
• Did it contradict earlier evidence?
• Is the gap important enough to spend interview time on?

Return:

1. Best next probe — one candidate-facing question.
2. Why now — one short interviewer-only line.
3. Optional second probe — only if the first answer could branch in two materially different directions.
4. Move on when — the evidence threshold for stopping.

Do not regenerate the full guide unless asked.

Round handoff mode

Read references/multi-round-interviews.md. Produce:

1. Evidence established
2. Evidence still weak
3. Contradictions to resolve
4. Do not repeat
5. Recommended next-round perspective
6. Top 3 next questions

Keep claims traceable to the notes. Use Not evidenced rather than assuming a negative result.

Quality check

Before responding, verify:

• every question maps to a material competency or risk;
• the guide fits the duration;
• sufficient prior questions are not repeated;
• candidate-facing text does not leak evaluation notes;
• follow-ups test ownership, scale, judgment, and outcomes;
• protected characteristics are excluded;
• strong and risk signals are balanced enough to limit confirmation bias;
• resume claims remain claims until verified;
• requested interviewer perspectives are visible;
• output is usable live, not merely comprehensive.

Adaptation rules

• First round: broad role fit, core capability, motivation, and major transfer risks.
• Second round: weak evidence, methodology, operating judgment, and execution constraints.
• Third/final round: unresolved contradictions, strategic trade-offs, leadership scope, and cross-functional credibility.
• Peer interview: collaboration contracts, conflict, handoffs, and credibility.
• Frontline interview: operating rhythm, execution, coaching, escalation, workload, and employee reality.
• Specialist interview: methodology, craft depth, decision quality, and measurable impact.

Failure modes

• Generating a generic question bank before reading the resume and job description
• Repeating a previous question because the answer was long rather than weak
• Revealing the desired answer inside the scenario
• Treating polished STAR structure as proof of impact
• Asking too many shallow questions for the available time
• Over-indexing on one stated concern and ignoring disconfirming evidence
• Scoring personality, accent, style, or similarity instead of job evidence
• Producing analysis without a copy-ready candidate-facing section
• Returning a full guide when the interviewer asked only for the next probe