TrialDiff Event-Class Evidence Records v0.1: a claim-bounded, hash-pinned record set for ClinicalTrials.gov protocol-amendment changes

100 deterministic, independently verifiable Evidence Records (across 52 trials) generated over a 100-trial ClinicalTrials.gov corpus of breast-cancer phase 2–3 interventional trials. Each record pins a single protocol-amendment change by content hash and carries a sorted set of definitional event-class memberships, explicit supported and non-supported claims, and an uncalibrated triage label (calibration_status: uncalibrated — not a validated review priority). The package includes the records, a SHA-256 manifest, and a validation note documenting the generation and export commands, rule-set hashes, and determinism evidence (byte-identical regeneration and re-export). Records assert only the structural fact of each change; they do not assert misconduct, regulatory non-compliance, manuscript non-disclosure, or clinical significance. An attempt to certify a review-priority severity label did not pass blinded review and is documented in the repository; the certifiable layer is definitional event-class membership.