# Adversarial release audit v1

**Date:** 26 October 2026

**Status:** passed

**Machine authority:** [audit JSON](adversarial_release_audit_v1.json)

## Result

A deterministic, state-stratified sample of **32** claims has zero source-digest or coverage failures. All **4** deliberate overclaim, type, provenance and propagation mutations are rejected. Six high-similarity, differently classified claim pairs have authored scope dispositions and no unresolved contradiction. Two external-blocker classes name the missing datum, in-scope impossibility, searched neighbourhood, affected branches, reopening trigger and review event.

This is adversarial evidence, not omniscience. It does not prove that every differently worded sentence in 6,000-plus files is mutually consistent. It makes the sample reproducible, attacks the controls directly and preserves the exact boundary of the claim.

## Mutation controls

- **MUT-OVERCLAIM:** Delete the non-implication coverage state from a sampled node. Rejected by claim graph completeness rule.
- **MUT-TYPE:** Replace a collision-safe claim identity with its bare legacy number. Rejected by path-plus-hash identity rule.
- **MUT-PROVENANCE:** Alter one byte of a sampled canonical source. Rejected by canonical source digest rule.
- **MUT-PROPAGATION:** Promote a reference-only relation into automatic transitive defeat. Rejected by explicit dependency-only propagation rule.

## Semantic candidate dispositions

- **CLM-1217-2145976f / CLM-1790-793e3271:** compatible_refinement — The earlier result reaches the pair-Gram carrier while explicitly retaining the orientation/affine need; the later result isolates that missing bit rather than denying the earlier carrier result.
- **CLM-0181-3f104f0d / CLM-0718-55856efa:** compatible_scope_restriction — Existence of a weak-basis invariant is compatible with the later claim that the relevant alignment obstruction needs such an invariant and appears only at high spurion order.
- **CLM-1644-cf35bdd3 / CLM-1673-9df7a90d:** compatible_nested_scope — Torsionless-background ultralocality and massive-branch second-class ultralocality are nested scope statements, not opposite verdicts.
- **CLM-0004-f22c24c5 / CLM-0005-12b3d5dc:** parallel_historical_surfaces — These are neighbouring synthesis surfaces with different declaration coverage; their state labels describe indexing status, not contrary scientific propositions.
- **CLM-0094-dbdefec5 / CLM-0106-b53dca1d:** compatible_sector_specialization — The custodial singlet Hessian is a sector specialization of the full-lock Hessian calculation, so the different graph signals do not express negation.
- **CLM-1709-57277955 / CLM-1712-0e68f16b:** distinct_component_channels — The fourth- and ninth-pole statements concern different flat-trace components; differing pole order is the claimed distinction, not a contradiction.