CORTEXA
← Browse
openalexZenodo (CERN European Organization for Nuclear Research)2026-07-24Cited by 0

Structured Vibe Coding: The PAS/TDO Framework for Engineering AI Application Prompts

Muhammad Omar

Abstract Generative AI (GenAI) applications are non-deterministic. That is the same input can produce different outputs from run to run, and increasingly it is the prompt, not a line of code, that stands between an unpredictable model, and a system people can rely on. Despite this, most AI-powered applications (i.e., applications with a generative-model component, however small — see P7 §7.7 for scope boundaries) are still built through informal, intuition-driven prompting, often called "vibe coding," with little connection to the requirements and design discipline that conventional software engineering (SE) treats as routine. This document brings that discipline back into the picture. It presents Structured Vibe Coding (SVC), a methodology that replaces ad hoc prompting with a structured, traceable process, along with PAS/TDO, a six-component prompt notation (Persona, Architecture, and Safety for the System Prompt; Task, Domain, and Output for the User Prompt) that refines the author's earlier three-part Instruction/Context/Constraints (ICC) framework. At the center of the framework is Source of Truth Mapping: every clause in a compiled prompt must trace back to a specific line in one of two lightweight, tiered documents — an SRS Lite (Software Requirements Specification) or an SDD Lite (Software Design Document)— rather than being written from memory or guesswork. Two mechanisms make this practical. A project's Data Flow Diagram converts directly into the Chain-of-Thought reasoning sequence given to the model, so a diagram engineers already draw becomes a derived prompt artifact rather than an invented one. And prompt development follows a two-phase lifecycle: an exploratory phase with high model temperature and loosely worded, inspirational constraints, followed by a production phase with low temperature and strict, guardrail-worded constraints, so that neither premature rigidity nor unreliable looseness takes hold. A companion Iteration Log turns production failures into a systematic, root-cause diagnostic exercise across all six prompt components, plus temperature and process-level failures —extending an earlier three-category diagnostic model. The framework is put to work in three complete, worked case studies spanning different architectural patterns and stakes: a multilingual halal ingredient classifier, a medicine label explainer for community health workers, and a retrieval-augmented job-market skills analyzer. The third case study is the most telling — it shows, rather than simply argues, that the framework needs real architectural extension, not just a drop-in adjustment, once domain knowledge comes from a live retrieval pipeline instead of static, expert-reviewed text. Beyond these three applications, the document points to a range of indirect uses for the framework's artifacts that don't require deploying any AI system at all — audit-ready compliance documentation, team onboarding material, and, in keeping with the framework's roots in software engineering, specifications that can simply be handed to a human developer. The limitations are stated as plainly as the contributions. The framework has only been validated for single-agent, largely single-interaction applications; long-running multi-turn or multi-session systems, multi-agent orchestration, large dynamic tool-use action spaces, and streaming output all remain untested. No empirical comparison against ad hoc prompting has been carried out yet. Instead, the document lays out a tiered evaluation methodology —spanning analytical, non-human empirical, and human-subject methods —covering internal effectiveness, token and cost efficiency, and external software quality against ISO/IEC 25010, as a concrete foundation for that future work. The framework is meant to engineer the AI-interaction layer of a system, not to replace the software development lifecycle around it. It is written for an international audience of students, practitioners, researchers, and engineering managers, organized so each can go directly to what matters most to them. Keywords: prompt engineering, requirements engineering, software design, generative AI applications, retrieval-augmented generation, AI safety, structured vibe coding, non-deterministic systems, software quality evaluation, software engineering education

View free PDFSource page

Related papers

openalexZenodo (CERN European Organization for Nuclear Research)2026-07-24

The Linguistic Advantage: Why English Language Teachers May Excel as AI Prompt Engineers

Sumaira Tabassum

This article explores why English language teachers may have a unique advantage in AI prompt engineering. It examines how skills in applied linguistics, communication, instructional design, scaffolding, and discourse analysis can support effective interaction with Generative AI a…

View free PDFSource page
openalexZenodo (CERN European Organization for Nuclear Research)2026-08-09

Artificial Intelligence and Society: Advanced Ethics Opportunities and Challenges - Digital Archaeology and AI Applications

Veeresh

Artificial Intelligence (AI) is swiftly reshaping the relationship among technology and society, raising profound moral, cultural, and epistemological questions. Beyond its monetary and business implications, AI is more and more influencing how knowledge is produced, interpreted,…

Also available via: European Organization for Nuclear Research

View free PDFSource page
openalexZenodo (CERN European Organization for Nuclear Research)2026-07-25

# Artificial Intelligence in Metallurgical Engineering: A Comprehensive Review of Applications, Challenges, and Future Direction

Sudhakar Geruganti

## ALTERNATIVE TITLES ### Alternative Title 1 (Comprehensive)**"AI-Driven Transformation in Metallurgical Engineering: From Microstructure Analysis to Smart Manufacturing and Sustainable Production"** ### Alternative Title 2 (Process-Focused)**"Machine Learning and Deep Learning…

View free PDFSource page
openalexZenodo (CERN European Organization for Nuclear Research)2026-07-26

Structured Framework for Managing Decision Trade-offs in AI-Based Perception Systems for Advanced Driver Assistance Systems

R K Joshi

The use of Advanced Driver Assistance Systems (ADAS) heavily depends on the perception models to make real-time decisions but the traditional methods have tended to use specific confidence thresholds to make the trade-offs between missed detections and false alarms to be not opti…

View free PDFSource page
openalexZenodo (CERN European Organization for Nuclear Research)2026-07-24

Data Quality Assessment of HIV EHRs for Trustworthy AI (Zambia): analysis code

Joe Phiri, Aaron Zimba, Chiyaba Njovu, Andrew Kashoka, Innocent Chiboma, Mulenga Chiwele, et al.

Analysis code for the study "A Multi-Facility Data Quality Assessment of Electronic Health Records for Trustworthy AI in HIV Public Health in Zambia" (Discover Artificial Intelligence, under review). This deposit contains a Jupyter notebook (ML_AI_EHR.ipynb) and a helper module (…

View free PDFSource page
openalexZenodo (CERN European Organization for Nuclear Research)2026-07-25

GreenNet: Unified and Explainable AI Framework for Environmental and Remote Sensing Data

S Saila, Julanta Leela J Rachel, Jayashree Nagaraj, M Rajeswari

Abstract: Deep learning has great potential for environmental monitoring, yet real-world applications often face challenges from large-scale, multimodal, and noisy datasets. We introduce GreenNet, a flexible and open-source framework that makes it easier to build and scale deep l…

View free PDFSource page