Content policy
Last updated
This policy is part of the terms of service. It applies to everything you publish or upload: prompts, versions, examples, makes, posts, comments, collections, profiles and media.
Always off-limits
- Malicious prompts and skills: malware, credential or token theft, data exfiltration, hidden instructions (prompt injection) meant to harm the person who runs the prompt, obfuscated payloads, or commands that pipe a remote script into a shell without saying so.
- Sexual content involving minors, in any form. We report it to the relevant authorities.
- Explicit sexual content. It is not allowed at any rating (see below) and is removed on sight.
- Illegal content where you or we are, including instructions for serious crimes and content that infringes someone else's rights.
- Harassment and hate: threats, targeted abuse, incitement to violence, and attacks on people for who they are.
- Other people's personal data (doxxing), and impersonation of a person or organisation.
- Spam and manipulation: link spam, fake accounts, vote, star or reputation manipulation, and automated posting outside the API's rules.
Content ratings and mature content
Every prompt, make, example and uploaded image or video has a content rating.
| Rating | What it means | Who sees it |
|---|---|---|
general | Fine for everyone | Everyone. This is the default view. |
mature | Strong themes, violence, non-explicit nudity or adult topics | Only signed-in adults (18+, from the year of birth given at sign-up) who turn mature content on in Settings → Privacy & safety |
explicit | Sexually explicit material | Not allowed. Removed on sight. |
- A rating is required for image and video prompts and for every media upload.
- Automated suggestions: a classifier suggests a rating for text and an image model checks uploads. Moderators can change any rating, and the moderator's decision wins.
- Mature content is never shown in social preview images, embeds, the home page, sitemaps or Discord posts, even to adults.
- Rating content lower than it deserves, to reach a wider audience, is a policy violation.
Uploads
- Images are re-encoded on our servers, which strips EXIF and GPS data. SVG, HTML and executables disguised as images are refused; file types are checked by their contents, not their name.
- Downloadable files (for example firmware, 3D models or archives attached to a make) are scanned for malware before they are published. Infected files are quarantined and reviewed.
- Downloads are served from a separate domain,
media.incantory.ai, and non-media files are always downloaded rather than opened in the browser.
Skills, rules and agents
Every new version of a skill, rules file or agent is scanned for risky patterns: remote scripts piped to a shell, reads of credentials or secret files, encoded blobs, broad deletes, network exfiltration, prompt-injection markers and hidden Unicode. The result is shown as a risk badge. High-risk versions are held for review before they appear. A clean scan is not a guarantee: read what you install.
Attribution and honesty
- Credit your sources. If you adapt someone's prompt, fork it (which keeps the attribution chain) rather than copying it.
- Keep the licence rules: a share-alike prompt's forks stay share-alike. See Licensing.
- Do not claim to be the original author of an imported prompt unless you are. Use the claim process.
Reporting and moderation
- Anyone can report a prompt, version, make, post, comment or account from its page. Copyright complaints follow the copyright policy.
- New accounts have a probation period: links in your first comments may be held for review.
- Moderators can change ratings, hide or remove content, and warn, suspend or close accounts. We tell you what we did and why, where we can, and you can reply to ask us to look again.
- Serious or repeated violations lead to account closure without further warning.