Greenhouse

// Build: internal ATS (Greenhouse replacement) # Paste into Cursor or Claude Code in an empty repo. Build an internal applicant tracker covering the part of Greenhouse we use: a candidate pipeline, interview scheduling, scorecards, a debrief synthesis, and an EEOC-grade audit log (this part is load-bearing). Stack: Next.js + Postgres + Calendar API + S3 (CVs) + Claude. Data model: - candidates(id, role_id, stage, source) - interviews(id, candidate_id, interviewer_id, scheduled_at, scorecard jsonb) - audit_log(id, actor, action, entity, at) // append-only, immutable Build, in order: 1. A pipeline board by stage; scorecards per role. 2. Interview scheduling across interviewers (free/busy + round-robin). 3. An append-only audit log with row-level immutability + retention policy. 4. AI assist, debrief synthesis. Call Claude: """ Given a candidate's full loop (scorecards + notes), the rubric, and the manager's must-haves, return JSON { per_area[], strengths[], risks[], recommendation(hire|no-hire|down-level|another-loop), disagreements[] }. Quote interviewers anonymously ("Interviewer A"). No protected-class inference. Don't write the offer. """ Skip: the nurture CRM, Sourcing/Marketing pillars, the non-SQL reporting. Done when: a loop runs end to end with a debrief read, and every action is in an immutable audit log.