Promptsto

Promptsto is a proposed open standard and platform for sharing prompts and templates used in natural language processing and generative AI. It envisions a centralized registry where users publish prompts along with structured metadata, including task category, model compatibility, input and output formats, placeholders, and any reported performance notes. The goal is to create a reusable, searchable collection of prompting approaches that supports reproducibility and cross-model experimentation.

Entries in promptsto typically include prompts or prompt templates, a description of the intended task, example prompts, and guidance on how to adapt the prompts to different contexts. Metadata fields may cover model constraints, input length, safety considerations, evaluation criteria, licensing, authorship, and version history. To facilitate interoperability, prompts are described using common schemas and can be stored in machine-readable formats such as JSON or YAML, with tags that enable precise filtering and discovery.

Governance and licensing are central to promptsto, emphasizing transparent provenance and attribution. Licensing terms often encourage sharing while protecting author rights, and moderation mechanisms address problematic content and safety concerns. Community governance may involve maintainers, contributors, and users who review prompts, propose improvements, and curate high-quality entries.

Applications of promptsto include benchmarking prompting strategies, supporting researchers in reproducing experiments, and aiding educators and practitioners in selecting effective prompts. By enabling standardized metadata and reusable templates, promptsto aims to reduce duplication of effort, enhance comparability across models, and foster collaboration in the prompt engineering community. Overall, it represents an aspirational framework for organizing and applying prompts in a scalable, responsible manner.