Screen every resume against the same rubric

Prompt 1 turns a job description into a screening rubric: three to five genuine deal-breakers phrased as yes/no questions, plus weighted criteria with a description of what strong, adequate, and weak evidence looks like. It refuses to invent requirements the JD never states, and it asks you about vague spots instead of guessing. You edit the result by hand before anything gets scored — the model drafts the rubric, you decide it.

Prompt 2 scores each resume independently against your final rubric. Every judgment needs a word-for-word quote from the resume as evidence; no quote means the score is zero and the field reads NOT FOUND. Bias guardrails run through all three prompts: names, gender, ages, graduation years, addresses, photos, school prestige, and employment gaps sit on an explicit ignore list, and the audit pass re-reads the scoring reasoning to flag anything that leaked in. The same prompts run in ChatGPT, Claude, or Gemini, so the pack works with whatever assistant your team already uses.

Run Prompt 1 on your job description in a fresh chat and edit the rubric it returns. That edit is the step that matters: cut criteria you would never reject someone over, fix the weights, tighten the must-haves.

Open a new chat for scoring. Paste the final rubric inside Prompt 2 with up to five resumes; small batches keep the scoring careful. Then run Prompt 3 in the same chat. It verifies every quoted piece of evidence actually appears in the resume, recalculates the weighted totals, checks the reasoning for ignore-list signals, and puts every UNCLEAR and every zero into a NEEDS HUMAN REVIEW list.

Before you trust any of it, calibrate: screen five resumes yourself for your first role, then compare. Where you and the rubric disagree, fix the rubric, not the score. And keep a human making every decision — a FAIL means someone takes a closer look, never an automatic rejection. Some jurisdictions regulate automated tools in hiring, so check the rules where you operate.

This recipe holds up when you hire for a role or two at a time. It stops being the right tool when applications arrive faster than anyone can paste them into a chat, when scores need to land in your ATS automatically, or when you need a consistent audit trail across every role and every hiring manager. At that point you want a system that reads each application the moment it arrives, scores it against your rubric, queues interviews, and keeps a human making every call. That is what we build — and a free AI opportunity audit will tell you whether your hiring volume justifies it yet.