A prompt library is only useful if you can trust it: trust that the prompts are safe, that they do what they say and that you are allowed to use them. This article explains the rules we follow for every prompt in AI Prompts Catalogue.
Only collections that allow reuse
People in the AI community publish thousands of prompts, skills and rules files, but not all of them may be copied. We only include collections released under a licence that allows anyone to reuse the content. At the moment that means four licences:
| Licence | What it allows | What it asks |
|---|---|---|
| CC0 (public domain) | Any use, without conditions | Nothing, although credit is appreciated |
| MIT | Any use, including commercial | Keep the copyright and licence notice |
| Apache 2.0 | Any use, including commercial | Keep the notices and say if you changed the work |
| CC BY 4.0 | Any use, including commercial | Credit the creator, link the licence and say if you changed the work |
If a collection has no licence, or a licence that forbids reuse, it is not in the catalogue, however popular it is. When a collection gathers items from many places, we check them one by one and only keep the items that carry a suitable licence.
What we leave out
Open does not mean anything goes. Before anything is published we remove:
- sexually explicit content and anything unsuitable for a general audience;
- “jailbreaks” and other attempts to get around an AI tool’s safety rules;
- offensive hacking tools and instructions that could cause real harm;
- duplicates, broken entries and one-off requests that are not reusable prompts.
For the example images that illustrate image prompts we go further. We leave out any picture that shows a real person, a copyrighted character, a brand logo, political imagery or a real tragedy, even when its licence would allow it.
How we tidy and sort
Raw collections are messy, so every prompt goes through the same clean-up:
- Titles and descriptions are cleaned so you can tell what a prompt does at a glance.
- Language and AI tools are detected, so you can filter for prompts in Spanish or prompts written for Cursor.
- Categories. Each prompt is placed in one of 13 categories and one focused collection inside it.
- Fill-in fields are recognised, so the parts you need to change become easy-to-fill blanks.
- Near-duplicates are merged. If two independent works share a name, the collection name is added to the title so you can tell them apart.
How credit works
When a licence asks for attribution, the prompt’s page says where it came from. Look for these details beside the prompt:
- Source: the collection, linked to its original repository;
- Prompt by: the author, when the collection names one;
- Licence: the licence name, with a link to its full text.
Example images carry their own credit and licence underneath the picture, with a note that we resized them. The Sources and licences page lists every collection with its full licence notice. If you reuse a prompt yourself, it is good practice to pass the same credit along.
Prompts from our members
Members can share their own prompts. Every new account is approved by a person, and every submission is reviewed before it is published. When members submit a prompt they confirm that it is their own work (or free to share) and release it under CC0, so anyone can use it without conditions. Their name appears on the prompt page.
Spotted a problem?
We check carefully, but with thousands of prompts something will slip through. Every prompt page has a Report this prompt link that tells us exactly which prompt you mean. If you believe something infringes your copyright, follow the steps on our Copyright and Takedown page and we will act quickly.
