Candidate Screener
Score applicants against an approved rubric with cited evidence, draft tailored interview questions, and get a ranked shortlist — never auto-answers a candidate.
This recipe makes screening defensible instead of vibes. It loads the role's rubric (must-haves, nice-to-haves, knockouts, weights) — drafting one for approval on first run — then evaluates every application in the window: per-criterion scores citing the exact resume line, knockout checks applied visibly before scoring, and every applicant banded shortlist / maybe / no-fit with a one-line rationale. Optional public-profile enrichment arrives as a cited appendix that never moves a rubric score. Per shortlisted candidate it drafts 4-6 interview questions probing the specific thin spots in their evidence. Interview invites and replies are staged drafts behind per-item approval — the recipe never auto-sends to a candidate, and reruns pick up only new applicants.
How it runs
Multi-agent orchestration — here's the flow, step by step.
Load the role's rubric from memory or draft one from the role description for user approval before any scoring happens.
job description writerPull application emails in the window by the role's alias/label or sender pattern, deduplicated against the screening ledger.
inbox scannerExtract resumes and cover notes from attachments and bodies and normalize per-applicant context.
email inbox managerCondense long resumes and portfolios into the evidence base each criterion is scored against.
document summarizerScore each application using the document-summarizer’s normalized evidence base against the approved rubric with the resume line cited per criterion, and apply knockout checks visibly before scoring.
recruitment screenerBand every applicant shortlist / maybe / no-fit with a one-line rationale grounded in cited evidence only.
triage routerWhen enrichment is enabled, add cited public-professional context as a separate appendix, never merged into the score.
web researcherRender the score distribution so the ranking is auditable rather than a black box.
data analystDraft 4-6 tailored interview questions per shortlisted candidate probing the specific thin spots in their evidence, plus a shared core set.
recruitment screenerDeliver the ranked shortlist, full roster, questions, enrichment appendix, and coverage footer in-hub.
report formatterStage interview-invite and reply drafts behind per-item approval; nothing candidate-facing sends automatically.
workflow orchestratorOn approval, stage approved invite drafts in Gmail and return receipts; hold everything in-hub otherwise.
distributorRequired Agents
11- recruitment-screener
- inbox-scanner
- email-inbox-manager
- document-summarizer
- job-description-writer
- triage-router
- data-analyst
- report-formatter
- web-researcher
- workflow-orchestrator
- distributor
Connections
Required
Optional
What it does
- Rubric approved by you before any candidate is scored — the yardstick is never invented
- Per-criterion scores citing the exact resume line or email evidence
- Knockout checks applied visibly before scoring, never folded into a vague no
- Full roster banded shortlist / maybe / no-fit with one-line rationales
- Tailored interview questions probing each candidate's specific evidence gaps
- Optional cited public-profile enrichment as a quarantined appendix
- Interview invites staged as drafts — nothing is ever auto-sent to a candidate
- Idempotent per role + window — reruns screen only new applicants
Example prompt
Screen this week's applicants for our Senior Bookkeeper role — score them against our rubric, rank a shortlist, and draft interview questions for the top ones.
Ready to deploy Candidate Screener?
Start free. One click, full agent orchestration.
Get Started Free →