Framework Document
The Van Vliet Prompting Playbook v1.1
A systems-oriented framework for extracting reliable, verifiable, and scalable outputs from probabilistic models.
Overview
The Van Vliet Prompting Playbook v1.1 is designed primarily for operators, researchers, builders, and teams, but it is written so that a general reader can begin with the Quick Start and adopt the advanced provisions gradually.
It is structured around prompting discipline, adversarial critique, deterministic validation, recursive improvement, and the design of fail-closed autonomous systems.
Document Scope
This playbook presents a practical method for producing more reliable work with generative AI. It begins with ordinary prompting (defining the task, context, output, and checks) and then escalates only when a chain, graph, or agent is justified. The framework is designed primarily for operators, researchers, builders, and teams, but it is written so that a general reader can begin with the Quick Start and adopt the advanced provisions gradually. Its governing principle is that the evaluation standard should be defined before generation.
The playbook emphasizes explicit success criteria, independent validation, bounded iteration, controlled tool use, provenance, and human responsibility.
Publication Note
This page serves as the public entry for the PDF edition. Future revisions may expand this page with supporting notes, examples, or implementation commentary.