what is code-to-prd fr?

code-to-prd is a skill for OpenAI's Codex that reads an existing codebase and turns it into a Product Requirements Document, or PRD. Instead of writing code, it works in reverse: you point it at a GitHub repository, a local checkout, a frontend app, a backend service, or a full stack project, and it produces documentation that describes what the product does, aimed at people who are not reading the source code directly. The skill scans the project's structure, the framework it uses, its routes and pages, its API endpoints, its data models, its enums, its permission rules, and how data flows between parts of the app. It also tries to tell the difference between real API calls and mock or fixture data, so the resulting document reflects what the software actually does rather than placeholder examples. It works across frontend, backend, and full stack codebases. The output is a folder of PRD documents: a system overview, page level docs describing each screen or view, an inventory of API endpoints, a dictionary of enum values used in the code, and a map of how pages relate to each other. The writing style aims to be readable by business stakeholders while still keeping the implementation details engineers or coding agents would need. This particular fork is set up for Codex users, with extra attention paid to Chinese language prompts, since the README shows example requests in Chinese such as asking the skill to break down a GitHub repo and generate a PRD in Chinese. It is installed through Codex's built in skill installer, pointing at this GitHub repository, and includes two plain Python scripts that can also be run directly: one that analyzes a codebase and produces a JSON or markdown summary, and one that scaffolds a PRD folder from that summary. The project is based on an existing MIT licensed code-to-prd skill from another author, credited in the README.