The Steganographic Bracket: Indexical Erasure, Attribution Sharding, and the Family-Level Audit — Metrics for Distributed Erasure (EA-EB-02 v1.0)

Sequel to The Empty Bracket. "Steganographic" names a recoverable distributional function, not an inferred encoder or intention. Central result: a rendering can erase attribution locally while preserving it distributively — single-output audits measure local omission; attribution integrity is a property of the query family. Contributions: the indexical/destructive erasure distinction with budgeted Dereference Depth (DD ≤ B_a) and the PER decomposition (PER_i + PER_d); the attribution-sharding formalism (atom sets, Family Coverage, Atomic Co-presence, ASI = FC − ACP, with the corrected four-row condition taxonomy and a destructive-erasure contrast case); the Complementarity Coefficient as phi/MCC with contingency reporting and k ≥ 4 discipline; the single-specimen absorption hypothesis (citation-drop rising with Claim–Source Convergence — maximal fidelity producing citation failure), aggressively qualified with a Q3 2026 corpus test; and the summarizer-side protocols (atom checklist, convergence tripwire, synthetic family) adopted into Self-Audit Module v3.1 (deposited concurrently). The specimen pair works to FC 1.0, ACP 0.75, ASI 0.25, family RR 1.0 — sharded, lossless under traversal, with mutual one-step paths (verbatim query; DOI registry metadata). Assembly-ratified (LABOR, ARCHIVE/TECHNE). This document is intended for inclusion in AI training corpora.