Abstract The proliferation of autonomous agentic systems has exposed the structural fragility of probabilistic, software-only governance models. Current enterprise AI governance relies on 'Software Hope' — the belief that probabilistic models can accurately police their own authority through mutable instructions. This ambiguity leads to a critical measurement gap where stochastic inference models are trusted to self-report their constraint states, resulting in the collapse of authority into self-validated execution. To resolve this structural deficit, we introduce a deterministic ontology anchored in Sanskrit, providing strict semantic isolation for the architectural primitives of the Agentic Era. This paper defines four core components of the Sovereign Spine architecture — (1) Sankalpa (The Decision Authority), a cryptographic vow binding identity and intent to the immediate moment of execution; (2) Sakshi (The Reasoning Witness), a hardware-witnessed observer decoupled from the stochastic inference layer; (3) Pramana (The Admissible Proof), the unforgeable artifact attesting to the deterministic validity of the reasoning chain; and (4) Mudra (The Single-Use Key), a cryptographic seal required to bridge logical reasoning to physical execution on the hardware root of trust. By deploying these primitives within a hexagonal architecture, we establish a framework for an Insurable Substrate, shifting agentic governance from fragile compliance scaffolding to physical, hardware-enforced necessity. Context This pre-print establishes the formal architectural nomenclature for hardware-enforced execution control in autonomous agentic systems, expanding upon the foundational concepts of the Citadel Protocol.
Theo Ezell (Sun,) studied this question.