Prompt Engineering in Multi-Agent Generative AI Systems: A Survey of Techniques, Design Patterns, and Open Challenges
Shweta Jain
, Anju Pawar , Nishchay Gupta
Prompt Engineering, Multi-Agent Systems, Large Language Models, Generative AI, Chain-of-Thought Reasoning, Retrieval-Augmented Generation, Agentic AI, Prompt Optimization .
Prompt engineering is the practice of directing large language model behaviour. Techniques developed for single-model inference zero-shot prompting, few-shot learning, and chain-of-thought reasoning are now being used inside multi-agent systems for which they were not designed. These systems are pipelines where models coordinate, delegate, and check each other's outputs. In this context, a prompt is not merely a request; it specifies an agent's role, capabilities, interactions with other agents, and behaviour under failure. This paper surveys more than 35 works from 2022 to 2026 and shows that prompt sensitivity propagates and amplifies in cascaded multi-agent architectures. This makes joint prompt-topology optimization a systems-level design problem rather than an isolated tuning task. We identify three open challenges that current production systems have yet to resolve.
"Prompt Engineering in Multi-Agent Generative AI Systems: A Survey of Techniques, Design Patterns, and Open Challenges", JETNR - JOURNAL OF EMERGING TRENDS AND NOVEL RESEARCH (www.JETNR.org), ISSN:2984-9276, Vol.4, Issue 6, page no.a171-a177, June-2026, Available :https://rjpn.org/JETNR/papers/JETNR2606027.pdf
Volume 4
Issue 6,
June-2026
Pages : a171-a177
Paper Reg. ID: JETNR_234937
Published Paper Id: JETNR2606027
Downloads: 000105
Research Area: Science and Technology
Country: Indore, Madhya Pradesh, India
ISSN: 2984-9276 | IMPACT FACTOR: 9.87 Calculated By Google Scholar | ESTD YEAR: 2023
An International Scholarly Open Access Journal, Peer-Reviewed, Refereed Journal Impact Factor 9.87 Calculate by Google Scholar and Semantic Scholar | AI-Powered Research Tool, Multidisciplinary, Monthly, Multilanguage Journal Indexing in All Major Database & Metadata, Citation Generator
Publisher: RJPN (IJPublication) Janvi Wave
