prompts.chat

Maintains dual persistence between a PostgreSQL database and a flat-file prompts.csv, enabling Git-based version control and contributor attribution tracking. The system uses a synchronization layer (scripts/generate-contributors.sh) that bidirectionally syncs changes, allowing community contributions via pull requests to automatically update the database while database mutations can be exported back to CSV for version control. This architecture enables both programmatic access (via API/database) and human-readable, version-controlled prompt management.

Provides a hierarchical discovery system supporting ChatGPT, Claude, Gemini, Llama, and other LLM models through category and tag-based filtering. The system uses a configuration-driven approach (prompts.config.ts) to define categories, tags, and discovery paths, with Server Components rendering filtered prompt lists (discovery-prompts.tsx) that support both curated and algorithmic discovery. The architecture separates content discovery logic from rendering, allowing different discovery strategies (homepage curation, category browsing, tag filtering) to coexist.

Supports importing and exporting prompts in multiple formats (CSV, JSON, YAML, etc.) with automatic format conversion and validation. The system can bulk-import prompts from external sources (e.g., GitHub repos, CSV files) and export the library for backup or migration. Import validation checks for required fields and data integrity, with error reporting for invalid records.

Enables creation of multi-step prompt workflows where the output of one prompt feeds into the next, with execution orchestration and state management across steps. The system supports conditional branching, loops, and error handling, allowing complex reasoning chains to be defined declaratively. Workflow state is persisted, enabling resumption and debugging of long-running chains.

Provides educational resources for learning prompt engineering, including an interactive prompt writing guide and a kids learning game that teaches prompt concepts through gamification. The system includes structured lessons, interactive exercises, and progress tracking, with content tailored to different skill levels (beginner to advanced). The kids game uses game mechanics (points, badges, levels) to make learning engaging.

Provides a command-line interface for managing prompts locally, including operations like search, create, edit, delete, and batch operations. The CLI can interact with both local files and remote instances (via API), enabling developers to manage prompts from their terminal without a web browser. The tool supports scripting and automation, with output formats suitable for piping to other tools (JSON, CSV).

Provides browser extensions (for Chrome, Firefox, Safari) and desktop applications that enable prompt access and execution from any web page or application. The extensions allow users to highlight text and apply prompts without leaving the current page, with context-aware prompt suggestions based on the selected text. Desktop apps provide native UI and offline access to the prompt library.

Exposes the prompt library as a native MCP server, allowing IDEs like Cursor and Claude Desktop to query and execute prompts directly from the editor without leaving the development environment. The MCP integration (referenced in README.md 137-148) provides tool definitions that map to prompt CRUD operations and discovery endpoints, enabling AI assistants to access, search, and apply prompts as part of their reasoning loop. This architecture treats the prompt library as a first-class tool in the MCP ecosystem rather than a web-only resource.

Provides one-click prompt execution through 'run buttons' that route prompts to multiple LLM providers (OpenAI, Anthropic, etc.) based on user configuration. The system stores provider API keys securely and implements a routing layer that maps prompt content to the appropriate provider's API format, handling model-specific parameter translation (e.g., temperature, max_tokens). This architecture decouples prompt definition from execution environment, allowing the same prompt to run on different models without modification.

Supports complex prompts composed of multiple files (SKILL.md structures) with automatic dependency resolution and composition. The system allows prompts to reference other prompts or skills, with a resolution layer that flattens dependencies into a single executable prompt. This enables prompt reuse and modular prompt engineering, where common patterns (e.g., 'think step-by-step', 'use XML output') can be defined once and composed into larger prompts.

Provides deep customization of branding, theming, and feature availability through a configuration-driven system (prompts.config.ts and feature flags). Organizations can self-host with custom logos, color schemes, domain names, and selectively enable/disable features (e.g., user authentication, collections, social features) without code changes. The configuration layer is evaluated at build time and runtime, allowing both static customization (branding) and dynamic feature toggling (feature flags).

Provides comprehensive i18n support for multiple languages and locales, including right-to-left (RTL) language rendering for Arabic, Hebrew, and other RTL scripts. The system uses a locale management layer that handles language selection, translation loading, and RTL CSS overrides. The architecture supports both static translation files and dynamic locale switching, with Server Components rendering locale-specific content server-side for SEO and performance.

Implements a flexible authentication system supporting multiple auth providers (OAuth, email/password, etc.) with role-based access control (RBAC) for users, moderators, and admins. The system manages user profiles, permissions, and content ownership, enabling users to create private collections and control who can view/edit their prompts. Authorization checks are enforced at both the API and UI levels, with role-based feature gating (e.g., only admins can moderate content).

Provides an admin interface for content moderation, user management, and bulk operations on prompts. The admin dashboard enables moderators to review flagged content, manage user accounts, and perform batch actions (e.g., bulk delete, bulk tag, bulk publish). The system includes content filtering and safety checks, with audit logs tracking all moderation actions for compliance and transparency.

Enables users to create and share collections of prompts, with a feed system that surfaces new prompts and collections from followed users or curated sources. The social features include following, liking, and commenting on prompts, with a recommendation algorithm that personalizes feed content based on user activity. Collections can be public or private, with sharing controls for collaborative curation.