ztn-process

Process new voice transcripts from _sources/inbox/ into ZTN three-layer architecture (Records + Knowledge + Hubs). Full pipeline: pre-scan with People Resolution Map, LLM noise gate, 16-question classification (incl. content potential, concepts, privacy trio), atomization, cross-domain scan, hub detection, per-batch subagent dispatch with self-review, people profile enrichment, idea living documents. Emits batch report per run and Evidence Trail initial entry in every new knowledge note.