Turn a role brief into a structured interview process you can defend. Map job outcomes to questions, calibrate anchored scorecards, preserve independent evidence, and hand the hiring owner a traceable recommendation.
Interviews go sideways when every panelist tests a different job, changes the bar mid-call, or writes a score with no evidence behind it. This skill starts with the approved role, hiring authority, candidate-data boundary, accessibility route, and the outcomes someone must demonstrate on the job.
It maps each competency to one observable claim, a consistent evidence method, a panel owner, and four role-specific score anchors. The workflow covers question and work-sample design, interviewer calibration, candidate instructions, factual note capture, independent scoring, contradiction review, panel debrief, and consistent recovery when a stage fails or the evidence is incomplete. Protected, private, pedigree-based, and unrelated information stays out of the score.
The handoff includes an interview contract, role evidence register, competency map, panel plan, question and exercise cards, calibration record, session cards, evidence logs, scorecards, debrief record, process-consistency review, open gates, and an evidence-matched state. It does not conduct a call without the required tools, invent candidate evidence, make employment decisions, or turn missing evidence into a low score.
Input
The job post lists twelve requirements, but the hiring manager says only service ownership, API design, production debugging, and cross-team handoff are must-haves. Four panelists each wrote their own questions. One wants a live coding puzzle with no tools; another wants to judge culture fit. There is no shared scoring bar.
Output
A role evidence register that routes the conflicting requirements, four observable competencies, a stage and panel map, standardized behavioral and work-sample cards, tool and accessibility rules, role-specific four-point anchors, excluded evidence, a synthetic calibration exercise, candidate instructions, and a plan-ready or blocked state tied to owner approval.
Provide the approved job description or role scorecard, level, hiring owner, interview stages, panel, timing, decision rule, candidate-data boundary, accessibility route, and any existing questions or notes. Ask the agent to build a new interview kit, prepare a panel, score supplied interview evidence, run a debrief, or repair an inconsistent process. Draft requirements, missing interviews, prohibited data, unverified notes, and owner decisions remain explicit instead of being filled with plausible content.
The role title and level, approved job description or scorecard, must-have and trainable outcomes, accountable hiring owner, decision rule, process state names, and applicable location or jurisdiction boundary.
Stages, time boxes, panel members, competency ownership, candidate instructions, allowed tools, ATS or handoff format, deadline, conflicts of interest, and any existing questions or rubric versions.
Only the candidate materials, transcripts, work samples, and notes authorized for this stage, plus privacy, retention, consent, recording, accessibility, prohibited-topic, and escalation rules. Missing evidence is allowed and stays visible.
SKILL.md; evidence, question-design, calibration, scoring, fairness, privacy, exception, and decision controls; reusable interview contract, role evidence, competency map, panel plan, question, work-sample, session, scorecard, debrief, reconciliation, and handoff templates; OpenAI interface metadata
No reviews yet.
Input
Three interviewers supplied notes for one candidate. One has detailed observations and a 3, another wrote only 'not senior enough' and a 1, and the third heard a medical disclosure before giving a 4 for resilience. The system-design exercise failed halfway through because the shared editor disconnected.
Output
Separate direct, reported, inferred, missing, contradicted, and prohibited evidence; the medical disclosure quarantined; unsupported scores held out; the failed exercise kept incomplete rather than failed; independent competency scorecards; a consistency-safe recovery option; critical open gates; and no decision-ready claim until the missing evidence is resolved.
Input
Six candidates interviewed for the same product manager role. The panel added commercial strategy after candidate three, gave later candidates an extra case prompt, and discussed scores before interviewers submitted notes. Two candidates received accommodations, but the tracker labels their longer sessions as unfair assistance.
Output
A preserved role and rubric version history, candidate-by-candidate process reconciliation, the legitimate format changes separated from bar changes, contaminated evidence flagged, affected competencies and candidates identified, a consistent remedy routed to the hiring owner, locked-scorecard and debrief controls, and a hold or no-decision state until the owner records the correction.
Trebuh
Updated Jul 13, 2026
Automated scan passed. It found 1 finding; highest severity info.
Scanned Aug 6, 2026.
Automated scanning reduces risk but cannot guarantee a package is safe for every environment.
Creator
Rrasvio
