SPG v2.2 (Super Process Guardrail): A Linguistic FSM Prompt Architecture for Deterministic and Stable Large Language Models in Production Environments

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Modern Large Language Models (LLMs) possess exceptional natural language processing capabilities but suffer from inherent limitations regarding non-determinism, programmatic drift, and token jitter. This paper introduces SPG v2.2 (Super Process Guardrail), a production-grade prompt architecture that implements a Linguistic Finite State Machine (L-FSM) directly inside the context window of an LLM. By combining static logic analysis (L.O.M Module), closed input scoping, and an output state-locking mechanism (Stability Engine), SPG v2.2 enables the transformation of probabilistic LLMs into stateful, deterministic execution engines, meeting the rigorous standards of Agentic Workflows.