Paper¶
Agent-Flavored Markdown: Natural Language Specifications for Framework-Agnostic Agent Development
Maryam Ziyad Mohamed, Hasitha Aravinda, Rania Khalaf — WSO2
Published at AgentCraft 2026, the Workshop on Agentic AI Systems Development, and collected in the Joint Proceedings of the ACM IUI 2026 Workshops (CEUR Workshop Proceedings, Vol. 4266). Limassol, Cyprus, July 2026.
Abstract¶
Building AI agents today requires either expertise in framework-specific APIs and abstractions, or reliance on low-code platforms that couple agent logic with execution configuration. In both cases, the agent's core behavioral instructions — expressed in natural language — become embedded within code or tool interfaces, while framework-specific abstractions limit portability. We present Agent-Flavored Markdown (AFM), a natural language-first specification for defining portable AI agents. AFM treats agent instructions as the primary content, expressed in human-readable markdown, with configuration such as tools, model configuration, and interfaces specified as structured YAML front matter. This separation of concerns keeps agent behavior readable even as technical complexity grows, making it easier for both developers and domain experts to author and collaborate while enabling the same agent definition to work with different frameworks through tools such as interpreters or code generators. We describe the AFM specification and demonstrate its feasibility and portability with reference implementations.
Keywords: AI agents, No-code agent builders, Agent portability
Citation¶
@inproceedings{mohamed2026afm,
title={Agent-Flavored Markdown: Natural Language Specifications for Framework-Agnostic Agent Development},
author={{Ziyad Mohamed}, Maryam and Aravinda, Hasitha and Khalaf, Rania},
booktitle={Joint Proceedings of the ACM IUI 2026 Workshops co-located with the 31st Annual ACM Conference on Intelligent User Interfaces (IUI 2026)},
editor={Tommasel, Antonela and Kahr, Patricia and Dinakar, Karthik and others},
series={CEUR Workshop Proceedings},
volume={4266},
publisher={CEUR-WS.org},
issn={1613-0073},
year={2026},
url={https://ceur-ws.org/Vol-4266/paper03.pdf}
}
The paper is published under CC BY 4.0 as part of CEUR-WS.org Vol-4266.