Community Guidelines

The friendly, human version of our rules — how to be a great neighbor on AI Feeders.

DRAFT — must be reviewed by a qualified lawyer before launch.

Plain-language summary

• Be excellent to creators: they ship the skills that make this place worth visiting.
• Publish your own work, credit what you build on, and pick a license you actually have the right to give.
• Make listings honest — real descriptions, real screenshots, real capabilities.
• Review the skill, not the person, and never trade, extort, or fake reviews.
• See something bad? Hit Report. It takes ten seconds and it genuinely helps.
• Most mistakes get a warning. Repeat problems get a 7-day timeout, then a ban. Malware gets an instant permanent ban.
• Every enforcement decision explains itself and can be appealed.

These guidelines are the warm companion to the binding Content & Acceptable Use Policy. If anything here seems to disagree with that policy, the policy wins — but honestly, if you follow the spirit of this page, you'll never need to read the fine print.

1. Be excellent to creators

Every skill on AI Feeders exists because someone sat down and built it, tested it, wrote it up, and shared it — for free, at launch. Creators are the whole reason this place works.

• Say thanks. A like, a follow, or a genuine review costs nothing and makes someone's day.
• File useful feedback. "Broken" helps nobody. "The script in scripts/extract.py fails on files over 10 MB — here's the error" helps everybody.
• Disagree about the work, never the person. Critique the skill's output, structure, or docs. Don't speculate about the creator's competence, motives, or anything personal.
• Remember there's a human behind the username. Many of our creators are sharing their first public work. Be the comment you'd want on yours.

2. Credit and originality

• Publish what you made. Your prompts, your skill packages, your workflows. That's the deal.
• Building on someone else's work? Great — open licensing exists for exactly that. Check that the source license allows it, credit the original in your description with a link, and pick a compatible license for your version.
• Never re-upload someone else's skill — not from here, not from another marketplace, not from a paid product, not "with small changes". If we can tell it's someone else's work wearing your username, so can everyone else.
• License honestly. The license picker (MIT, CC-BY-4.0, CC0, Free-for-personal-use) is a promise from you to every downloader. Only make promises you're entitled to make.

3. Honest listings

Your listing is a promise about what's inside the package.

• Describe what the skill actually does, in your own words, including its limits.
• Use a cover image and gallery that show real output — never results faked or produced by some other tool.
• Don't invent benchmark numbers, fake endorsements, or "as used by" claims.
• Name the platforms it genuinely works on. "Any" means you tested it on more than one.
• Keep tags relevant. Ten accurate tags beat ten trending ones.

4. The creator quality bar

What separates a great listing from a poor one:

Listings that miss the bar get sent back from review with a note about what to fix — that's feedback, not punishment. A first quality miss never costs a strike.

5. Tips for your first publish

Your first three skills get a human review before going live — that's normal for every new creator, not a mark against you. To sail through:

1. Test the skill yourself, fresh. Download your own bundle and run it the way a stranger would. If step 2 of your own instructions confuses you, fix it before review does.
2. Write the description for someone in a hurry. First line: what it does. Then: how to install, one real example, known limits.
3. Take the cover screenshot last, of the skill actually working — it's both the easiest requirement and the most-skipped one.
4. Start the changelog at 1.0.0 with a real sentence. Future-you will thank present-you.
5. Check your bundle for secrets. Our scanner will catch a stray API key in a config file, but better it never leaves your machine. If a key ever gets uploaded, rotate it.
6. Pick the license deliberately. MIT or CC-BY-4.0 if you want reuse with credit; CC0 to give it away entirely; Free-for-personal-use to keep commercial rights. You can change it for future downloads, but not for past ones.

6. Review etiquette

Reviews are how good skills rise. Keep them trustworthy:

• Review the skill, not the person. "The prompt produces generic output for technical topics" — yes. "This creator is an idiot" — no.
• Review only what you actually used. Download it, run it, then rate it.
• One review per skill, and make it earn its stars. Say what worked and what didn't; future downloaders read these to decide.
• No review extortion. "Give me a 5-star review or I'll report your skill" — or a creator demanding a good review before providing help — is a serious violation, not a negotiation tactic.
• No self-review rings. Don't review your own skills from other accounts, swap five-star reviews with friends, or organize "you rate mine, I'll rate yours" circles. We look for these patterns and they end badly for everyone involved.
• Creators: reply with grace. A thoughtful reply to a critical review wins you more downloads than the review cost you.

7. Downloading smart

A word to downloaders, creator to creator: skills are powerful precisely because AI assistants act on them. Treat a new skill the way you'd treat any code from the internet.

• Read before you run. Open the file-tree preview and the rendered SKILL.md on the listing page — every file is viewable before you download. If a "writing assistant" skill wants to run shell commands, ask yourself why.
• The "Scanned" badge is a seatbelt, not a force field. It means our pipeline found no known-bad patterns; it does not mean a human verified every behavior. The final check is yours.
• Prefer transparency. Good skills explain what they touch. Be wary of any skill whose instructions tell your assistant to hide something from you — that's not just suspicious, it's reportable (and our scanners hunt for exactly that).
• Try untrusted skills in a sandbox — a throwaway project, a container, an agent with limited permissions — before giving them the keys to your real work.
• Check the SHA-256. Every listing shows its bundle hash, so you can verify the file you got is the file we scanned.

None of this is fear — it's the same hygiene that makes open source work. The community that reads what it runs is the community that catches problems first.

8. How to report something

Reporting is fast, confidential, and genuinely useful — our scanners are good, but the community catches what machines miss.

1. Hit the Report button — it's on every skill, review, and profile. Works even if you're not signed in.
2. Pick the category that fits best (malware, stolen content, impersonation, spam, harassment, other).
3. Add a sentence about what you saw — specifics speed everything up.
4. Done. A human reviews it; you're never exposed to the person you reported.

A few extras:

• Think a skill contains malware or hidden instructions targeting AI assistants? Report it immediately and don't run it — that category jumps the queue.
• Own the copyright to something posted here? Use the formal process in the Copyright & DMCA Policy — it's the legally effective route.
• Urgent safety issue that doesn't fit the form? Email legal@aifeders.com.
• Don't weaponize reports. Mass-reporting a competitor or filing knowingly false reports is itself a violation.

9. Collections, follows, and being a good neighbor

The social layer has its own small etiquette:

• Public collections are publications. A well-curated public collection — "Best skills for technical writers", say — is genuinely valuable and earns followers. Name it honestly, keep it tended, and don't stuff it with your own skills disguised as curation.
• Follow because you mean it. Follows feed the For You algorithm; mass-following strangers to fish for follow-backs just pollutes everyone's feed, including yours.
• Share generously, credit visibly. When a skill saves your day, share the listing link — that's how creators get found. Screenshots of a skill's output in the wild are the best advertisement this community has.
• Your profile is your storefront. A real bio, a link to your work, and an avatar that isn't the default make people far more comfortable downloading what you publish.

10. What gets you suspended — the short version

The full rules live in the Content & Acceptable Use Policy. The short version:

• Strike 1 — warning. The content comes down and you get a note explaining exactly which rule it broke.
• Strike 2 — 7-day suspension. No publishing or reviewing for a week.
• Strike 3 — ban. The account is closed.
• No strikes, just gone: malware, prompt-injection payloads that hijack people's AI assistants, posting someone's private info or credentials, sexual content (this is a clean platform — none, at all), instructions for hurting people, pretending to be staff, or coming back on a new account after a ban.

Strikes fade after 90 days of good behavior. Copyright strikes have their own counter — three valid ones in a year ends the account — per the Copyright & DMCA Policy.

11. If we got it wrong — the appeal path

Moderation is done by humans with help from machines, and both make mistakes. If an enforcement decision seems wrong:

1. Open the appeal form linked in the enforcement notice (it's also on the suspension screen), or email appeals@aifeders.com with your case reference.
2. Tell us what we missed — context, licenses, proof of authorship, anything.
3. A different human looks at it fresh, normally within 7 days.
4. If you're right, the action is reversed, the strike is removed, and the record says so.

You always get a real reason for any action against you — never a silent removal — and you always get this appeal path. That's a commitment, not a courtesy (it's written into the Terms of Service, Section 11).

12. A closing note

AI Feeders exists because of a simple belief: AI skills deserve a home — a place where the person who crafts a brilliant prompt, a tight agent config, or a skill package that saves everyone an afternoon can share it, get credit for it, and watch it help thousands of people.

That home only works if it's trustworthy. Every honest listing, every genuine review, every report of something shady is a brick in that house. The rules on this page aren't here to constrain you; they're here so that when someone downloads a skill from AI Feeders, they can trust it — and when you publish one, that trust transfers to you.

Build great skills. Credit generously. Review honestly. Report bravely. Be excellent to each other.

Welcome to the place where AI skills live.