How jujumilk3/leaked-system-prompts works
Main Components
• Repository Layer: At the highest level is the Git repository. Represent the version control system as a component managing all files.
• Index/README: The README file acts as an entry point or index that links to each leaked prompt.
• Prompt Files: Each markdown file (with specific names including model names, versions, and timestamps) can be represented as individual components or grouped by category (e.g., Anthropic Claude prompts, OpenAI prompts, etc.).
• Images/Assets: The “images” directory is a dedicated folder, if needed, for any graphical resources related to the collection.
Architectural Patterns and Design Principles
• Flat File Organization: The structure relies on a flat file system with meaningful naming conventions (model name, version, timestamp) rather than a complex codebase.
• Metadata as File Naming: The system leverages file names to encode metadata (like timestamps, model types, etc.). Represent this by annotating the prompt file components with their metadata.
Relevant Technologies or Themes
• Git: Emphasize the use of Git for version control.
• Date/Versioning: Indicate that file names include both versioning and time-stamped data.
• Start at the top with a “Repository” node that encapsulates the version-controlled project.
• Branch out to a “README” node that serves as the landing or navigation point.
• Branch from the README node into clusters or groups based on prompt file categories (e.g., group by language model or by timestamp/version).
• Metadata such as model names and versioning to capture the prompt file characteristics.
• Using annotations or legends to explain the significance of the names/timestamps in context.