One claim per line. Optional evidence follows with double pipes: claim || date || jurisdiction || authority || primary HTTPS URL.
No claim has been treated as verified.
Compromise extracts language. It does not verify truth, law, or current status.
All three statements remain HOLD until claim-specific provenance is attached and inspected.
Input lines enter the NLP gate, which unpacks raw claim text into separate components: extracted action verbs, places, and piped provenance metadata. The parsing gate validates five checks: date format, jurisdiction, authority, HTTPS primary record, and compound scope. A sample claim moves across each filter barrier. Any unverified line is held with explicit remediation. Once full evidence is provided, the claim routes to READY for manifest export.