TrialDiff Event-Class Evidence Records v0.1.2: a claim-bounded, hash-pinned record set for ClinicalTrials.gov protocol-amendment changes

Correction notice (Erratum E4): This immutable v0.1.2 deposit remains byte-verifiable, but its secondary-outcome predicate omitted three secondary_outcome_removed_after_primary_completion memberships because repeated JSON Patch array removals were resolved against the original array and whole-array removals were not recognized. No published membership was retracted; the correction adds three. It is corrected by v0.1.3. The concept DOI 10.5281/zenodo.20801956 always resolves to the current version. Deposited files remain unchanged. Corrected release (Erratum E1): v0.1.2 supersedes v0.1.1 for corrected event-class membership claims. It removes nine unsupported why_stopped_removed_in_terminal_context memberships caused by missing TO-version snapshots being interpreted as field absence. The shared TO-view derivation now reconstructs absent TO snapshots from adjacent-version patches. The same correction also captures two previously missed, patch-evidenced enrollment_changed_to_zero memberships. The immutable v0.1.1 deposit remains available with its erratum.This package contains 97 deterministic, independently verifiable Evidence Records across 54 trials, with 106 event-class memberships. Class counts are: enrollment_changed_to_zero 3; outcome_edit_cooccurs_with_results_posting 80; primary_endpoint_changed_after_primary_completion_without_results_reconciliation 10; secondary_outcome_removed_after_primary_completion 9; why_stopped_removed_in_terminal_context 4. The overlap distribution is 88 records with one class and 9 with two classes. The recomputed boundary analysis is 4,485 patches scanned, 73 inclusive primary-endpoint flags, 63 reconciliation-confounded flags, and 10 clean class members.The archive contains records/, docs/, MANIFEST.sha256, VALIDATION.md, and expected_stats.json. The docs include the determinism attestation, erratum, Evidence Record primitive specification, and the severity calibration/decoupling reports. Release ZIP SHA-256: 4681fb0e5baaab53fb9352721a49aaa7b5e2027a18c9029547592ff7dfb709e7. Event-class rule-set hash: 07957f8b90549d4f42387f51b471ecde9901b6db63bbc27b84c73631603407c0. Combined rule-set hash: 318445b9ad266f51fd10ef378645c753ba7a098e3e4395c3c457750dc5f88d86. Generator code tree: 6176b2121bdcd1cac6ce859d7749ab188de4b183. Frozen package commit and release-tag target: 8777d04c11e7e660a22db51d3589498911e7d086.Records assert only the structural fact of each change; they do not assert misconduct, regulatory non-compliance, manuscript non-disclosure, harmlessness, or clinical significance. Triage labels are uncalibrated metadata, not validated review-priority findings. Hashes and manifests establish byte integrity; semantic corrections are documented through errata and new rule-set generations.GitHub release: https://github.com/AMBRA7592/trialdiff/releases/tag/event-class-v0.1.2