{
  "schema": "csoai.coverage-register/0.1",
  "as_of": "2026-08-26",
  "audit": "Counts from the in-repo record (attestation-corpus.json, harness/rwa-attest TARGETS, interop runs). Nothing invented; unverified claims marked.",
  "ruling_ref": "BOARD-RULING.md (2026-08-26, corrected): card index frozen at the verifiable floor 150 until the 185 candidate hashes verify against the real card store; judged UNMEASURED before measured.",
  "chains": {
    "xrpl": {
      "instruments_in_coverage_index": 16,
      "located_issuers_with_measured_control_facts": 6,
      "declared_not_located": 10,
      "note": "16 XRPL entries in the coverage corpus (6 measured control facts via financial-measure-run-v2; 10 declared UNMEASURED/not-located, incl. Aviva/DCP/EURCV/JMWH pending public address location)."
    },
    "evm_eas": {
      "contracts_with_measured_facts": 3,
      "instruments_in_coverage_index": 2,
      "rails": "EAS off-chain attestation support; contract-level facts measured for BUIDL/BENJI/ACRED",
      "note": "'100 ETH instruments' is NOT in the record — aspirational (UNMEASURED); pipeline exists, coverage grows per located contract. Never counted as present."
    }
  },
  "grammar": "The board's size is not frozen here: cite live totals.public_count on GET /api/gspc, or totals.count_grammar for the long form. Two numbers travel together — totals.axes counts SLOTS, totals.measured_axes counts slots with a real run behind them, totals.unmeasured_axes is published so the gap is visible. Quoting the slot count alone claims measurements that do not exist; quote both, or quote the smaller.",
  "grammar_correction_2026-08-26": "This field previously read 'Public quotable board = 14 axes ... Never publish 22 axes as the board.' That was accurate while the 8 financial/domain axes were ruled in (2026-08-24) but absent from the signed board payload — a guardrail against typing a number the artifact could not back, not a repeal of the ruling. The 2026-08-26 sweep wired them into the signed data and re-signed it, so the endpoint now backs the larger count. The retired guardrail is recorded rather than quietly overwritten. See council-os/ADR-001-axis-count.md."
}
