A credible library, built for people who study prompts.
PromptBase collects the real system prompts behind shipping AI products, annotates what makes them work, and teaches the craft of writing your own.
Most “prompt libraries” are a wall of recycled one-liners with no provenance. PromptBase is the opposite. Every entry in the curated corpus is a genuine, published prompt — the kind that ships inside tools like Cursor, Claude Code, v0, Perplexity and Devin — reproduced with its source linked and its design choices explained.
What's inside
- 100 curated prompts spanning 43+ tools, each tagged with the techniques it uses and the things worth learning from it.
- A 19-lesson curriculum grounded in empirical prompt-engineering research, not folklore.
- A real community layer — create an account, publish your own prompts, save a collection, and discuss what works.
How we keep it credible
Three rules govern the library:
- Provenance over volume. We'd rather have one prompt with a clear source than a hundred anonymous ones. The curated corpus carries explicit attribution and licensing.
- Annotation, not decoration. Each prompt explains its notable design decisions so you learn the why, not just the what.
- Evidence in the curriculum. The Learn section is honest about what the research supports — including the popular techniques that don't hold up.
Attribution & licensing
The seeded corpus is reproduced from the open system-prompts-and-models-of-ai-tools collection under the GPL-3.0 license. Original authorship of each prompt belongs to the respective vendor. See the full sources & credits page. Community-submitted prompts remain the property of their authors.
If you find an attribution error or want a prompt removed, reach out and we'll correct it promptly.