# Annotation plan

- Downstream decision and risk:
- Task/unit/population:
- Label schema, abstention, and boundary examples:
- Guide/version owner and change process:
- Data source, rights/consent, sensitivity, retention/deletion:
- Splits: training / validation / held-out audit / challenge:
- Sampling mix and active selector:
- Annotator qualifications, pay/workload, accessibility, and escalation:
- Calibration, blinding, agreement, adjudication, and re-review:
- Cost-quality stopping rule and reopen triggers:
- Handoff to data engineering and ML engineering: