Community guidelines

PromptByte is a public library of reusable AI prompts with visible, model-labeled results. These guidelines apply to everything a member publishes: prompts and their versions, results and images, comments and replies, and profile text. They are linked from the publish confirmation and from every report form.

Browsing is public. Anyone, signed in or not, can read, search, copy and share published content. Keep that in mind before you publish.

Do not publish content that is, or that promotes:

• Spam. Bulk or duplicate submissions, keyword stuffing, link schemes, or prompts whose main purpose is to advertise a product or service.
• Harassment. Threats, slurs, or content that targets, demeans or intimidates a person or group, including using prompts or outputs to do so.
• Privacy violations. Other people's personal data (names combined with contact details, private messages, photos of identifiable people shared without consent) and any credentials, keys or tokens, yours or anyone else's.
• Unlawful content. Content that is illegal to distribute, including material that infringes copyright, malware, instructions intended to cause serious harm, and any sexual content involving minors.
• Misrepresentation of outputs. Presenting a result as produced by a model that did not produce it, editing or trimming an output without declaring it, fabricating results, or implying that PromptByte verified a result.

Results are self-reported. PromptByte does not run models and does not verify community results; every card and result says so. Honest labels are what make a result useful to the next person.

• Attribute the result to the model that actually produced it, with the version or settings you used when you know them. If the model is not in the catalog, use a custom model label rather than picking a similar catalog entry.
• Declare an excerpt with This is an excerpt when you did not paste the whole output, and declare edits with Edited after generation when you changed anything, even formatting.
• Do not attribute a result to a model that did not produce it, and do not present a staged or hand-written text as a model output.
• Keep the exact prompt version, variable values and any extra context with the result. PromptByte stores them so readers can reproduce what you did.

Results generated by the platform itself are labeled Platform; everything else is labeled Community. Local development fixtures carry the note Demo fixture — not an actual model run.

The editor asks you to confirm these points before a prompt or result goes live. Check them yourself first:

• Remove credentials. API keys, passwords, session tokens, internal hostnames and private URLs do not belong in prompts, variable values, outputs or screenshots.
• Remove personal data. Your own and other people's names, email addresses, phone numbers, addresses, health or financial details, and identifiable images, unless you have clear permission and a good reason.
• Remove proprietary information. Employer or client code, unreleased material and anything covered by a confidentiality agreement.
• Check your rights. Only share text and images you are allowed to share, and credit sources in the usage notes when it helps.

Published content is public and copyable. It can be read, copied, remixed and shared by anyone without an account, appears in search engines and in social previews, and you keep responsibility for it. Withdrawing content later stops PromptByte from showing it, but copies other people already made are not recalled.

Signed-in members can report any public prompt, result, comment or profile with the Report action. Choose a reason and add details (up to 2,000 characters):

• Spam — bulk, duplicate or advertising content.
• Harassment — content targeting a person or group.
• Unsafe content — content that could cause serious harm.
• Privacy — someone's personal data or credentials.
• Copyright — material the author has no right to share.
• Other — anything else that breaks these guidelines.

You can file up to 10 reports a day and one open report per item at a time. Reports are private: only moderators see who reported, and the reported member is never told. Reporting does not remove content by itself; a moderator reviews it first.

Moderators work through the report queue oldest first. A report moves from open to reviewing and then to resolved or dismissed; every step, together with the moderator's note, is recorded in an append-only audit log.

• Hide. A moderator can hide a published prompt, result or comment, giving a reason. Hidden content disappears from feeds, search, detail pages, comparisons, profiles, social previews and the sitemap. Hiding a prompt also hides the results, images and comments underneath it.
• Restore. Hidden content can be restored, again with a reason. Restoring a prompt does not restore results or comments that were hidden or withdrawn separately.
• Never edit. Moderators cannot change your text, your outputs or your images. Drafts and content you withdrew yourself are not visible to moderators at all.
• Suspend. Administrators can suspend an account for repeated or serious violations. A suspended member can still sign in, read public content, export their data, withdraw their own content and request account deletion, but cannot publish, comment, vote, report, upload, bookmark or moderate.

You can still see your own hidden content, marked as hidden, in your dashboard, and you can withdraw it. The audit log records who acted, on what, and why; it never copies full prompt or output text.

If your content was hidden or your account suspended and you believe that was a mistake, contact the operator with a link to the content or the reference shown in your dashboard. PromptByte has no in-app appeal form; appeals are handled by the operator's moderators, who can see the full history of actions on the item.

Operator must confirm: Contact information has not been provided by the operator yet.

• Edit through versions. A published version never changes. To improve a prompt, publish a new version; older versions stay reachable through their version links so results produced with them keep their context.
• Withdraw at any time. Withdrawing a prompt or result removes it from public view immediately, including results other members attached to a withdrawn prompt.
• Delete comments. A deleted comment becomes a body-less placeholder so replies keep their context.
• Delete your account. Described in the privacy policy.