The problem
Prompts get copy-pasted across repos and drift out of sync, with no history of what changed or why.
How it works
Every prompt lives in one versioned JSON store with rich metadata. A lightweight web UI makes the library searchable and editable for non-engineers.
Highlights
- Single source of truth for prompts across projects, preventing drift and duplication.
- Structured schema carries metadata per prompt: name, category, version, target model and usage notes.
- Web UI for browsing, searching and editing without touching raw JSON.
- Two parallel implementations: a v1 library and a v2 with extended features.