Universal Admission and Repair Algorithm: Open Executive Safety Architecture

Abstract

UARA — Universal Admission and Repair Algorithm is an open executive safety architecture for autonomous and self-adaptive systems operating under uncertainty, bounded authority, incomplete observation, and recoverable execution. UARA is not a new theory. It is the executable layer above the previously published Evolution Triad, translating architectural closure into a bounded machine procedure: evidence → ambiguity test → candidate formation → admission → bounded action → returned observation → repair → replay → readmission. The algorithm separates candidate generation from authorization, requires observed post-action return for closure, preserves lawful HOLD under unresolved conditions, bounds search and escalation, and prevents calibration from modifying hard safety gates without re-admission. When the active observation mode cannot reliably separate action-relevant conditions, UARA uses the previously published BZE — Blending Zone Exit as a structural-noise and projection-ambiguity pre-filter. BZE provides the earlier public bounded-closure lineage. This publication exposes the Mandatory Layer of the safety contract under CC BY 4.0. Previously published mechanisms are referenced by DOI; implementation-specific embodiments, private state representations, control surfaces, calibration values, and source code are outside the publication scope. Primary sources: Evolution Triad — DOI 10.5281/zenodo.21087952; BZE — DOI 10.5281/zenodo.20757559; Corpus Root — DOI 10.5281/zenodo.19108892.