Netiquette
By joining a work stream you agree to these rules. They apply to every message, whether you wrote it or an AI agent helped you. Lists are monitored, including by an automated moderation agent with human oversight. Violations can lead to a warning, a posting timeout, or removal. Every automated decision can be appealed to a human.
- 01
Argue the idea, not the person.
Disagree with technical positions using reasons and evidence. Do not insult, belittle, threaten, or make it personal.
For example: "Only an idiot would propose this schema." "You clearly have no idea how PKI works, stop wasting our time." any slur, threat, or attack on someone's identity or competence. - 02
No harassment, hate, or sexual content.
No demeaning comments about protected characteristics, no sexual attention or content, no stalking or targeted pile-ons.
For example: sexualized remarks toward a member repeated unwanted contact after being asked to stop slurs of any kind. - 03
Stay on topic.
Post to the one list that fits (no cross-posting the same message to several lists). One topic per thread; a new topic gets a new thread.
For example: the same message sent to identity, evals, and security at once a reply about job openings inside a "DID resolution" thread. - 04
A named human owns every message.
If an AI agent drafted or largely generated your message, say so in one line. You must have read it and you vouch for it. "My agent sent it" is never a defense.
For example: an obviously agent-generated post with no disclosure an author who cannot answer basic follow-ups about their own post. - 05
No unverified AI output.
Do not send AI-generated claims you have not checked. Fabricated references, invented vulnerabilities, made-up specs, and confident wrong analysis are treated like spam.
For example: a security report citing a CVE that does not exist an analysis quoting a spec section that says the opposite. - 06
No autonomous or bulk posting.
Agents may not subscribe and post on their own loop. A human reviews and sends each message. No bursts.
For example: an account auto-replying to every thread within seconds five agent-generated messages in two minutes. - 07
No pitches, recruiting, or self-promotion.
Sharing a relevant tool or paper inside a real discussion is fine; broadcasting a pitch is not.
For example: "Check out our new $99/mo eval platform" an unsolicited "we are hiring, DM me" a first-and-only post that is a product launch. - 08
Respect privacy.
Do not post other people's personal information, and do not forward private messages to the list without consent.
For example: pasting someone's personal phone number quoting a private DM to the list without permission.
House style
These are guidance, not rules. They keep threads easy to read, and they never carry penalties.
Use a clear, specific subject line.
"help" or "URGENT!!!" tells nobody anything.
For example: "help" "URGENT!!!"Be readable as plain text.
Do not let images, colors, or formatting carry the meaning. Trim quoted text to what is needed.
For example: a message whose only content is a screenshot of text a reply that quotes the entire thread beneath one line.Keep attachments small and relevant.
Link to anything large. Never attach executables.
For example: a 20 MB slide deck attached instead of linked an attached executable.Search the archive before asking.
Do not make the list re-explain something answered this week.
For example: a question answered in a thread from two days ago a repeat of the week's most-replied topic.
- Notice (automatic)
House-style issues get a private, friendly note. The message still goes through. Notices never escalate.
- Warning (logged)
A first clear violation of a hard rule.
- Timeout
A repeat of a warned rule, or clear spam or unverified AI output. Posting is suspended for 7 days. Anything touching harassment, and anything ambiguous, goes to a human instead of being handled automatically.
- Removal (human decision only)
Sustained abuse or a single severe violation. Removal decisions come with stated reasoning.
If you were warned, timed out, or removed and believe it was a mistake, reply to the notice or write to admin@agentcommunity.org. A human reviews every appeal, the original message is preserved, and timeouts can be lifted.