talent-ops

Evidence-based hiring operating system for AI coding CLIs. The employer-side mirror of career-ops: candidates got AI to choose companies — this gives hiring teams AI to choose candidates on evidence, not keywords.

Resume != Candidate. AI recommends, humans decide. Every decision has a reason code. Git history is your audit trail.

The local board — every applicant by stage, scored and evidence-backed. npm run board

How it works

Talent-Ops runs entirely on files inside your repo. AI CLI modes generate and score; you decide. Each step writes a small markdown file with YAML frontmatter, so the whole hiring history lives in git — diffable, reviewable, auditable. The pipeline, end to end:

• Define the role — define-role runs a calibration conversation

(business need, first-90-day outcomes, must-haves with the evidence that would prove them, disqualifiers, scoring weights) and writes a role-contract.md. Nothing downstream runs until you approve it.

• Generate the JD — jd turns the approved contract into a job

description, scrubbing biased language and refusing any requirement that isn't in the contract. A candidate-facing AI-use disclosure is appended.

• Intake — drop CVs (PDF/DOCX/TXT/MD, or a CSV) into data/inbox/;

intake parses, normalizes, dedupes and hard-filter-prechecks each into a candidate folder. Unparseable files go to quarantine.md, never silently dropped.