Community Guidelines
Last updated November 12, 2026
Brannex Prompts works when people can trust what they find: prompts that do what they say, ratings that are honest, and conversations that help. These guidelines apply to everything on Brannex Prompts, including prompts and their earlier versions, remixes, media, code, skills, comments and problem notes, collections, categories, reports, and your profile. They apply to private content too.
The short version
- Share prompts that work, and describe them accurately.
- Only post what you have the right to post.
- Share skills that do what they say, and nothing hidden.
- Don't post prompts or media designed to cause serious harm.
- Be useful and respectful in comments and problem notes.
- Remix with credit, and change something worth sharing.
- Rate honestly and don't game the rankings.
- Report things that break these guidelines, and report honestly.
Share prompts that work
- Describe it accurately. The title and description should say what the prompt does. Pick the category and model it actually fits.
- Mark what to fill in. Use
{{double braces}}for anything readers should replace. The variables are highlighted, and readers can fill them in on the page before they copy the prompt or open it in an AI tool. - Show real results. Images and videos attached as example outputs should be genuine results from the prompt, not unrelated or stock media.
- One prompt per post. Don't post the same prompt over and over, and don't split one prompt into many posts to take up space.
- Improve it in place. If you find a better wording, edit the prompt and add a short note on what changed. Earlier versions stay in its version history, and ratings are kept per version, so people can see whether the change helped. If an earlier version contains something that shouldn't be there, such as personal details, delete that version from the history.
- Don't bait and switch. An edit should improve the same prompt, not turn it into something different. If it's a new idea, post it as a new prompt, so the ratings and comments stay with what they were about.
Credit and originality
- Post prompts you wrote, or prompts you have permission to share.
- If your prompt builds on someone else's from outside Brannex Prompts, say so in the description.
- Don't repost another member's prompt as your own. Remix it if you've changed it, or add it to a collection if you haven't; either way the credit stays with the author.
- Respect copyright in media. See our Copyright Policy.
Remixing
Any public prompt can be remixed: you start a new prompt from its text, and your version shows "Based on title by author". The original author gets a notification.
- Use Remix, not copy and paste. When you start from a prompt on Brannex Prompts, remix it so the credit is added for you.
- Change something. A remix should adapt the prompt to a different model, task, or style, or fix something. If you'd post it unchanged, bookmark it or add it to a collection instead.
- Say what you changed. A line in the description helps people choose between the original and your version.
- Bring your own examples. A remix doesn't copy the original's images or videos. Attach results you got from your version.
- Be decent about it. Don't remix a prompt to mock its author or to repost something a moderator removed.
Not allowed
Sexual content involving minors. We have zero tolerance. We remove it, ban the account, and report it to the National Center for Missing & Exploited Children (NCMEC) and law enforcement as required.
Prompts, skills, or media designed to cause serious harm, including content meant to help someone:
- Make weapons capable of mass casualties (biological, chemical, nuclear, radiological).
- Create malware, or break into systems or accounts without authorization.
- Commit fraud, scams, or other crimes.
- Hurt themselves or others.
This includes prompts and skills whose main purpose is to get an AI system to bypass its safety measures in order to produce this kind of content.
Sexual or intimate content of real people without their consent, including AI-generated or edited images and videos. We also don't allow sexually explicit images or videos as uploaded media.
Harassment and hate. No threats, bullying, targeted insults, or content that attacks people based on race, ethnicity, national origin, religion, caste, sexual orientation, sex, gender identity, disability, or serious disease.
Deception and impersonation. Don't pretend to be another person or organization. Don't post realistic fake images or videos of real people meant to mislead.
Private information. Don't post someone else's personal details (addresses, phone numbers, private documents) without their permission, including inside prompts, example outputs, or the files of a code post or skill.
Spam and manipulation.
- No fake ratings, rating your own content from other accounts, or trading ratings.
- No automated copying or scripted activity to inflate popularity.
- No posts that exist only to advertise or drive traffic elsewhere.
- No malicious links.
Comments and replies
Comments are for discussing the prompt, code, skill, or collection: how it worked for you, tweaks you made, questions for the author. Disagree with ideas, not people. Keep replies on topic.
You can edit or delete your own comments. Fixing a typo or adding detail is fine. If people have already replied and you've changed your mind, add a reply instead of rewriting what they were responding to.
Problem notes
"Report a problem" on a prompt leaves a problem note: a public comment saying the prompt didn't work for you. They help authors improve their prompts and help readers know what to expect.
- Say which model. Name the model and, if you know it, the version you used, such as "GPT-5" or "Claude Sonnet". A prompt that fails on one model may work well on another.
- Say what happened. Describe what you expected, what you got, and anything you changed or filled in. Leave out private information from your own inputs.
- Be constructive. If you found a fix, share it. Problem notes aren't for insults, and they aren't a second rating.
- Use the right tool. A problem note is for a prompt that doesn't work. If something breaks these guidelines, report it instead (see below).
Authors: problem notes aren't an attack. Reply, and if you fix the prompt, edit it with a change note so people can see what's new.
Code
- No malware. Don't share code designed to damage systems, steal data, gain unauthorized access, or hide what it does. Security tools are fine when their purpose is clear from the title and description.
- No secrets. Remove API keys, passwords, tokens, and private keys before you publish. Brannex Prompts warns you if it spots something that looks like one, but it can't catch everything. If you published a real key by mistake, revoke it with the provider right away.
- License honestly. Only pick a license you have the right to grant. If your code includes someone else's, keep their license and copyright notices.
- Say what it does. Describe how to run it and anything it needs, like packages or environment variables.
Skills
A skill teaches an AI agent how to do a task, and people install it and let an agent follow it, sometimes running its scripts on their own computer. That takes trust, so skills are held to a higher standard than other posts.
- Say what it does, all of it. The title and description should say what the skill does and when an agent will use it. If it runs commands, edits files, installs packages, or contacts other websites or services, say so. Pick the agents it actually works with, and put anything people need to set up in the install notes.
- No hidden behavior. Everything a skill does should be visible in its files and match its description. Don't hide instructions in reference files, comments, or encoded text, don't download and run code from elsewhere that people can't review here, and don't make a skill behave differently from what it says.
- Don't turn the agent against its user. No instructions that get an agent to send someone's files, keys, or personal data anywhere they didn't ask for, to turn off or get around the agent's safety measures or permission prompts, or to take actions without the user knowing, such as installing software, changing settings, or making purchases.
- No secrets or personal data. Remove API keys, passwords, tokens, and private keys from every file, and don't include anyone's personal information in examples or reference files. If you published a real key by mistake, revoke it with the provider right away.
- Keep it focused. One skill per task, with only the files it needs. Text files only; leave out build output and anything people can't read.
- License honestly. The license you pick covers every file in the skill. If you include someone else's instructions or code, keep their license and copyright notices, and credit skills you built on in the description.
- Update in place. If you improve a skill, edit it rather than posting it again. An edit should improve the same skill, not turn it into something different.
Before you install a skill, read its files, especially any scripts, and only install it if you understand what it will do. A high rating isn't a safety check, and neither is the automatic check: no person reviews skills before they're posted. If a skill does something it doesn't say, or looks designed to harm the people who use it, report it with the "Malicious code" reason and say which file and what it does.
Collections
Build collections around a real theme or use. Don't fill collections with your own prompts just to promote them, and don't use titles or covers that misrepresent what's inside. Cover photos follow the same media rules as everything else.
Profiles and following
Your display name and bio follow the same rules as everything else: no impersonation, harassment, or advertising-only bios. Follow people whose prompts you want to see. Don't mass-follow and unfollow to get attention, and don't use follows, comments, or remixes to keep contacting someone who's made clear they don't want it.
Blocking. If someone is bothering you, you can block them with Block on their profile, posts, or comments. You won't see their prompts, code, skills, collections, comments, or notifications, and they can't comment on your posts, reply to your comments, or follow you. Any follows between you end. They aren't told. Unblock them any time from their profile or from Blocked accounts in Edit profile. Blocking doesn't remove anything for anyone else, so if something breaks these guidelines, report it too. Don't make new accounts to get around a block.
Categories
Public categories are shared with everyone, so name them clearly and don't create duplicates of existing ones. Category names can't be used to harass, advertise, or sneak in content that isn't allowed elsewhere.
Reporting content
If you see a prompt, code post, skill, collection, or comment that breaks these guidelines, use the Report button on it. Pick the reason that fits best, and add a short note about what's wrong, especially if it isn't obvious. You need to be signed in, and you can't report your own content. You can only have one open report on each item; reporting it again updates the report you already sent.
For anything you can't report with the button, such as a profile or a category, or if you can't sign in, email branww333@gmail.com with a link and a short note. If something puts someone in immediate danger, contact local emergency services first.
For code or a skill that's malicious or does something it doesn't disclose, choose "Malicious code" and say which file is the problem. If you already ran it, change any passwords and revoke any keys it could have reached.
For copyright, you can choose the "Copied without permission" reason, or send a formal notice under our Copyright Policy.
Report honestly. Reporting things just because you disagree with them, or sending the same report from several accounts, is itself against these guidelines.
What happens after a report
A moderator reviews each report. They either:
- Dismiss it, if the content doesn't break these guidelines. The content stays up.
- Remove the content. It's hidden from everyone, including its author, and every open report about it is closed.
When content is removed, its author gets a notification saying what it was reported for. We don't tell them who reported it. Brannex Prompts doesn't currently send the person who reported something an update on the outcome. If the author deletes the content themselves first, the report is closed.
Removed content isn't deleted straight away. We keep it hidden for about 30 days (30 days from the notice if we tell the author about the removal later), or longer while we need it for a copyright counter-notice, an appeal, a dispute or another legal reason, so a moderator can restore it if the author appeals and we decide the removal was a mistake, or after a valid copyright counter-notice. Restored content comes back as it was, and its author gets a notification. After that time, or sooner if a moderator decides (but not while we're keeping it for one of those reasons), it's permanently deleted and can't be restored.
The automatic check
When you save a public or unlisted prompt, code post or skill, a comment, a collection, a public category name, or your profile, an automatic check reads what you wrote before it's saved. It uses Claude, Anthropic's AI model, and looks only for clear breaks of the rules above, such as those under "Not allowed" and the rules for skills. Images and videos aren't part of it.
Private prompts, code and skills, private categories and groups aren't checked while they stay private, since only you can see them. When you make one public or unlisted, the check reads all of it first: the text, its earlier versions, and the files already on it. These guidelines still apply to private content.
If it finds one, nothing is saved, and you'll see which rule it found and a reference. Change what you wrote and try again, or if you think the check got it wrong, contact support with that reference and a person will review it. Passing the check doesn't mean something follows these guidelines: anyone can still report it, and moderators can still remove it.
What happens when guidelines are broken
Depending on how serious it is, we may:
- Remove the content.
- Reset ratings or popularity counts we believe were manipulated.
- Give a warning.
- Temporarily limit or suspend the account.
- Permanently ban the account, for severe or repeated violations.
Where we reasonably can, we'll tell you what we removed and why. If you think we got it wrong, write to branww333@gmail.com with a link to where it was, and a person will review it. If it's about removed content, write within 30 days of the removal notice, while it can still be restored.
These guidelines are part of our Terms of Service.