dot-prompt

dot-prompt provides a structured approach to composing, reviewing, and managing prompts for AI applications. It addresses the challenge of prompt invisibility—where developers and teams lose track of the exact instructions sent to large language models—by turning prompts into inspectable, reusable components.

prompt files, which can include branching logic, variables, and reusable fragments. This structure allows prompts to be separated from application code, making them readable and accessible to developers, product managers, designers, and domain experts alike. dot-prompt enables real-time preview of compiled prompts: users can select variables, see the exact output that will be sent to the AI, and instantly observe changes as variables are updated. Branching logic is resolved before submission, ensuring the AI receives a single, clear instruction rather than embedded conditionals.

The tool supports fragment composition, allowing prompt authors to build prompts from reusable pieces such as skill definitions, context blocks, and examples, with static fragments cached and dynamic fragments fetched afresh each time. Input and output contracts can be declared, and dot-prompt validates both sides to detect breaking changes automatically. Version control is built in, recognizing the difference between breaking and non-breaking changes and prompting for versioning when necessary, so older versions remain available for callers who have not upgraded.

dot-prompt is delivered as a language and viewer for writing and inspecting prompts, with files managed separately from codebases. It is intended for developers, product teams, domain experts, and founders who need visibility and control over AI prompt workflows.