Engineering Statements: A Structured Specification Framework for Agent-Oriented Scientific Workflows

Engineering Statements is a portable specification framework aimed at bridging the gap between scientific papers and engineering implementation. It transforms scientific intent into reproducible next steps via YAML-formatted structured declarations. Its core philosophy is: scientific papers describe discoveries, while this framework specifies how to inspect, disseminate, and extend these discoveries. The goal is to keep scientific intent verifiable, reproducible, and portable, while making next steps more accessible. The project is maintained by thinkthoughts, with the original source in the GitHub repository (link: https://github.com/thinkthoughts/engineering-statements).

The traditional scientific research process has obvious gaps: in the paper phase, researchers publish findings and methods → in the understanding phase, other researchers try to comprehend → in the reproduction phase, a lot of trial and error is needed → in the application phase, methods are applied to new problems. There is information loss and friction between each phase. Engineering Statements attempts to solve this problem by introducing a structured intermediate layer.

In terms of design philosophy, the framework clearly aims not to replace papers, repositories, or researchers, but to create an engineering-friendly specification layer. Its workflow is: Paper/Source → Engineering Statement (YAML) → Notebook → Experiment Report → Infographic → Published Product. The repository structure includes core components: statements/ (YAML specification files describing objectives, context, evidence, etc.), notebooks/ (demonstration notebooks), src/ (Python tool library supporting loading and validation), templates/ (reusable templates), outputs/ (generated products).

Practical application cases include number theory (e.g., research on arXiv papers Split primes and modular Nahim sums, generating a complete product chain) and AI safety (cross-model scale lie detector evaluation, ensuring reproducibility and auditability). In addition, this framework complements the AGAPI agent system, serving as a bridge from scientific infrastructure to reproducible actions, supporting AI agents to read specifications and execute experiments. Related ecosystem projects include Climate Reality (climate data application) and Lab Reports (experiment report tool).

This framework can alleviate the reproducibility crisis in the scientific community (providing machine-readable specifications to lower the threshold for reproduction); support agent-assisted scientific research (structured specifications become an interface for human-machine collaboration); and promote interdisciplinary knowledge transfer (unified specification format facilitates understanding and cooperation across different fields).

Engineering Statements represents the evolution direction of scientific research infrastructure: from static papers to dynamic specifications, from human reading to human-machine collaboration, from describing discoveries to specifying actions. Its value lies not only in technical implementation but also in a mindset shift—researchers think about how to make their work easy for machines to understand and execute. In the future, it is expected to form an interoperable scientific research agent ecosystem to address complex scientific challenges.