Agentic Publication Protocol: An Attempt to Modernize Scientific Publication
06:00 · June 29, 2026 · arXiv cs.AI RSS
Scientific publication is still organized primarily around static manuscripts, even though much of scientific progress depends on tacit know-how: how to run code, reproduce figures, interpret edge cases, choose useful follow-up directions, and avoid failed paths. Large language model agents create an opportunity to publish not only knowledge, but also operational know-how in a form that future readers and researchers can directly use. This paper outlines the Agentic Publication Protocol (APP), a lightweight repository format for packaging a paper together with code, data, environment information, reproducibility instructions, and an agent-facing instruction file. APP treats a version-controlled repository as the publication object and uses \texttt{AGENTS.md} and optional skills to define a paper agent that can explain the work, reproduce key results when possible, and support follow-up research. We describe the design principles and details of the protocol, as well as the agent skills useful for publishing papers under the protocol. We also describe development tools for evaluating and improving the protocol and associated agent skills. Finally, we provide a broader discussion of the future of scientific research in the agent era.
Summary
Scientific publication continues to rely on static manuscripts that convey conclusions but rarely preserve the operational knowledge needed to run code, reproduce figures, interpret edge cases, or identify productive next steps. The Agentic Publication Protocol (APP) proposes a lightweight alternative that treats a version-controlled repository as the primary publication object. In addition to the manuscript, an APP repository includes code, data, environment specifications, reproducibility instructions, and an agent-facing file named AGENTS.md. This structure allows an LLM agent to act as an interactive representative of the work.
The protocol is built around four design principles. Information is organized modularly so that both humans and agents can locate authoritative files quickly. Reproducibility is supported by explicit commands and scripts wherever feasible. Version control and release tags establish a precise, timestamped record of the claimed artifact. Finally, the AGENTS.md file and optional skills encode tacit know-how that static text cannot capture, such as which numerical settings matter or which attempted directions proved unproductive.
An APP-compliant repository can be published as a GitHub release, after which readers or other agents interact with a paper-specific agent to obtain explanations, attempt reproductions, or explore follow-up directions. The authors supply development tools and example skills to help researchers adopt the format, while noting that domain-specific extensions will likely be required. The approach aims to reduce the friction that currently slows verification and reuse, without replacing traditional peer review or credit mechanisms.
Why it matters
This research is highly relevant for Dutch AI researchers and academic institutions as it provides a practical framework to enhance the reproducibility and transparency of AI models. Aligning with the EU's push for trustworthy AI, adopting such agentic protocols could position the Netherlands at the forefront of modern, verifiable scientific publishing.
