Moderation
Punishments, reports, the chat filter, standing, chargebacks and blocked commands.
Reports #
Reports (staff Reports, permissions reports.view and reports.manage):
- In-game
/report:- Players with a portal account get a clickable link (via 1.8
tellraw) to/account/reports/new, already filled in with the player, category and server. There they can add detail and evidence links. - Players without an account report in chat with
/report <player> <category> <what happened>, or/report bug <what happened>. Staff see these tagged Unregistered user.
- Players with a portal account get a clickable link (via 1.8
- Categories: cheating, chat, griefing, exploit, bug (no player) and other.
- Limits: a 30-second cooldown, at most 5 open reports per player, and no reporting yourself.
- Captured automatically when a report is filed: the reported player's last 10 messages (from the past 30 minutes), whether they were online and on which server, and the reporter's location.
- Staff page:
- Filter by status, category, registered or unregistered reporter, player, or "assigned to me".
- Claim a report, set its status, and record an outcome note (required to resolve or close).
- A Punish shortcut, plus links to other reports about the same player.
- When a report is resolved or closed, the reporter is told in-game. Registered reporters can also follow status under My reports, and new reports go to Discord webhooks subscribed to
reports.
Chat filter #
Chat filter (/chat-filter, permission chatfilter.manage):
- Staff add words in bulk (one per line). Each rule matches as a whole word, anywhere in the message, or as a regex, and either blocks the message or censors the match with
****. - Word list files ("Upload word lists"): upload text files with one word or phrase per line, several at once, e.g. the raw files from LDNOOBW's list (
en,de,ja...). The file name gives the language (jabecomes the panel'sJP,zh→ZH,ko→KO,fil→TL), or staff pick one. A language's words are only sent to servers while that language is turned on under Translations; turning one on or off there updates servers within a minute. Japanese, Chinese, Korean and Thai words match anywhere, since those languages don't put spaces between words. Lines starting with#and duplicates are skipped. "Word lists by language" shows the count per language, whether it's on, and deletes a language's words (for re-uploading a newer list). - Whitelist: add words under Whitelist on the page, or choose "Whitelist it" when adding a word. A whitelisted word is never blocked or censored, even if it is on the filter itself or a filtered word is inside it. For example, whitelist
classicwhenassmatches anywhere. A respelling of a whitelisted word is still filtered. Only that word is skipped, so anything else in the message is still checked. "Whole word" allows just that word, and "Anywhere" also covers longer words (assassincoversassassins), but it can let through a bad word made by joining letters to it. Allowed words needChatFilter.skupdated on the server, and they aren't sent as rules, so older scripts ignore them. - A tester shows exactly what the game would do with a message.
- The panel converts each rule into a regex that works the same in PHP and Java.
- Servers fetch
GET /server/chat-filter?version=…every 60 seconds (chat_filter_interval). When nothing has changed, only{changed:false}comes back. - Servers compile the patterns once and match locally, with no API call per message. The last list is saved, so filtering keeps working after a restart or while the panel is down.
- Block rules also catch look-alike spellings (
0→o 1→i 3→e 4→a 5→s 7→t 8→b @→a $→s !→i, colour codes removed). Private messages containing a filtered word are blocked. - Respellings: a rule that is a single word of 4 or more letters (whole word or anywhere) also catches extra letters and gaps between every letter, so
peniscatchespennis,peeenis,p e n i sandp.e.n.i.s. A letter doubled in the rule stays doubled (buttdoesn't catchbut). Shorter words, phrases, regex rules and allowed words match exactly as written, and if a real word gets caught, add it as an allowed word. It catches variants of listed words only: it doesn't learn new words, and a different letter (penus) needs its own rule. - Blocked messages appear in the player's message log marked Blocked, with the original text.
network.chatfilter.bypassexempts staff, and/chatfilter [reload]shows rule counts or forces a re-download.
Turning chat moderation off for a category or server #
The Chat moderation rule is on everywhere by default. Turn it off in a category's rules (Servers → Groups → edit), or for one server in that server's rules, where "Same as group" follows the category. A server's own setting wins over its category's. Servers pick up the change with their group settings on the next heartbeat (within about 15 seconds).
- Off skips the automatic chat checks on that server (
network/Panel/Events.sk,network/Chat/ChatFilter.sk): the word filter (block and censor) in chat and private messages, the spam limits (cooldown, repeats, bursts, shouting, stretched letters) and the blank-message check. Messages go through as typed, still with colours and textmojis for players allowed them. - Manual moderation keeps working: mutes, bans, kicks, warnings and reports. Mutes already in place still apply, so a muted player still can't chat or send private messages there. Chat is still logged to the message log.
- Fails safe: until a server has its rules from the panel (just after a restart, or while the panel can't be reached), it moderates as normal. A server whose scripts are older than this rule ignores it and keeps moderating.
- The panel itself doesn't filter messages coming from servers, so nothing else changes there. The chat filter page lists the categories and servers where the rule is off.
Chat text and spam #
The Chat text and spam box on the chat filter page sets spam limits and what chat can show. The settings travel with the word list (settings in GET /server/chat-filter), so saving them updates every server within a minute, and the last copy keeps working while the panel is down. Changes are in the audit log (chatfilter.settings).
-
Spam limits (
network/Chat/ChatFilter.sk, checked after the word filter; 0 turns one off): seconds between messages, the same message again within a time (capitals, spaces and symbols ignored), and the most messages in a row within a few seconds. A stopped message gets a short note back and isn't logged as a punishment. Shouting (more than a set share of capitals, in messages with enough letters) is lowered, and long runs of one letter are cut (heyyyyyy→heyyyyat 4). Staff withnetwork.chatfilter.bypassornetwork.chatspam.bypassskip them. -
Premium players skip spam limits (on by default). The word filter still applies to them. While it's on, the store's perks table shows it.
-
Hover a name in chat shows the player's rank, level, badge, crew, Premium and kills, and clicking it types
/msg <name>. Game modes can add lines ({-hover::<uuid>::<key>}). Only on 1.8 (Pit) for now: on modern servers (Survival, Lobby) chat goes out the normal way, because a copy re-sent with hovers would skip the chat filter and mutes there. Kill message hovers work everywhere. -
Show items in chat:
[item]or[i]in a message becomes the item in the player's hand; others hover it to see its name, enchantments and lore. -
Show items in chat with
<item>(off by default):<item>works like[item]. If the player's hand is empty,<item>stays in the message as typed. -
Hovers in kill messages: names in kill messages get the same hover, and the killer's weapon is shown in brackets with its own hover.
-
Gradient text: who may use
<gradient:#ff5555:#5555ff>Hello</gradient>(two or more colours) and&#ff8800in chat: everyone, Premium players and staff (the default; shown on the store), staff only (network.chat.gradient), or nobody. Others see the words without the colours. Text from the panel (messages, item names, translations, broadcasts) can always use them. Pit (1.8) has only 16 chat colours, so each letter gets the closest one. -
Colour codes (
network/Chat/ChatText.sk): who may type&colour and formatting codes like&c(red) and&l(bold) in chat: everyone, Premium players and staff (the default), staff only (network.chat.colour), or nobody. Codes they can use limits which ones (0123456789abcdeflmnorby default: every colour and format except&k, scrambled text). Ranks that can always use them (comma separated, for examplevip, mvp) let those ranks use codes, gradients and hex colours whatever the setting. With Players can type the chat colours they own (on by default), a player may also type the colours they have from their rank, the store or staff, for example&bfor someone who owns aqua; the panel sends these as letters (chat_colours, e.g."7b") in the login profile and in profile updates. Everyone else has codes,&#rrggbb,<#rrggbb>and gradients taken out of their message, so only the words are left. A§typed by a modded client is treated as&. Game text from the panel and scripts is not affected. -
Stop blank messages (on by default,
network/Chat/ChatFilter.sk): a message with nothing to read in it is stopped with a short note, for example one that is only colour codes, spaces, or invisible characters (zero-width spaces, Hangul fillers, the blank Braille cell, combining marks). Invisible characters are also taken out of every other message, so they can't hide a word from the filter. This runs for everyone, staff included.
Players decide what others see about them: Show my stats when people hover my name and Show my weapon in kill messages (both on, under Privacy in /settings and on the portal). With stats hidden, the hover shows only their name.
How the hovers are sent: when a line needs one, the server takes the message's recipients off the chat event and sends the line with tellraw at least 2 ticks (100 ms) later, after every other script (mutes, filter, quest answers, the chat format) has had its say, so a cancelled message is never shown. Chat runs on its own thread, so the very next tick could come too soon; and if the format is somehow still the game's plain <name> message, the network's chat line is used instead. The console logs chat as normal. Kill messages work the same way at the end of the death event, after custom death messages.
Clickable text #
Links and buttons in chat are sent as JSON with both the old keys (clickEvent, hoverEvent, read by 1.8 to 1.21.4) and the new 1.21.5+ ones (click_event with url or command, hover_event). Several scripts (report and vote links, NetSafe_SendLink used by crew upgrades, custom command links and friend notices) only wrote the old keys, so they clicked on Pit and not on Paper; they now all go through Text_Part in core/Support/TextFormat.sk. Plain https:// and www. links in announcements are clickable too, and panel text can use <link:url|hover>label</link>, <run:/command|hover>label</run>, <suggest:/command|hover>label</suggest> and <hover:text>label</hover> in announcements and messages sent with Text_Send.
Combat logging #
Combat logging (/combat-log, permission combatlog.manage). Each server group sets whether combat logging is punished, the tag length, whether to broadcast it, and what the player loses: killed (items drop unless the group keeps inventory), inventory deleted, in-game money (coins::), Pit XP (cxp::), vanilla levels, any other name::%uuid% variable, global coins, or a console command. Amounts are flat (100) or a share (10%). Every combat log is recorded with who they were fighting, the time left on the tag and exactly what was taken, on the Combat logging page and the player's Combat logs tab. Players who logged 3+ times in a week are flagged.
Account standing and appeals #
Account standing and appeals. Every account sits on a five-point scale: Suspended, Danger, Warning, Good, Amazing, shown as a card on the portal (Account standing) and on the player's Punishments tab for staff. An active ban is always Suspended. Otherwise staff's latest decision counts (Set standing, permission standing.manage, with a comment the player sees). With no decision it follows the last 90 days of the record: nothing is Good, any warning, kick or mute is Warning, and an active mute or three or more is Danger. Players can appeal a punishment or a standing decision from the same page, once each. Staff answer under Appeals (/punishments/appeals, permission appeals.manage): reversing lifts a ban or mute in-game, or stops a warning or decision counting, and the player sees the reply. Staff roles that already exist need these two permissions ticked by hand.
Chargebacks #
Chargebacks. When PayPal reverses a store payment (a chargeback or dispute: IPN status Reversed, webhook PAYMENT.SALE.REVERSED), on top of taking back what it gave, the player is: flagged (a red Chargeback tag on their player page), banned for good ("Store (chargeback)"), their portal account suspended (not for staff accounts, so nobody locks staff out), and the public IP addresses they played from in the last 30 days IP banned. Staff get a red Discord message on webhooks ticked for store purchases, and an audit log entry. A refund staff give (Refunded) is not a chargeback. Undo any part by hand: revoke the ban on the player page, turn the account back on under Players → Accounts, and lift IP bans under Moderation → IP bans (permission punishments.view to see, punishments.ban to add or lift). IP bans refuse anyone joining from that address, whatever account they use; staff accounts are let through. Servers send the joining address with their login check (core/Players/Sessions.sk), so this needs the updated core scripts.
Blocked commands #
Blocked commands (network/Moderation/CommandBlock.sk): each category (Server groups → edit) has a list of commands players can't run, and each server (Servers → edit) can add more. One per line without the /: pl also blocks /bukkit:pl; gamemode creative blocks just that (and anything after it); worldedit:* blocks every /worldedit:… form; *:* blocks every /plugin:command form. Players get "That command is blocked here." Staff with the in-game permission neutine.commands.bypass aren't blocked, and the panel's console isn't affected. Changes reach servers with the next heartbeat. Commands still show in tab completion.
Every punishment goes through the panel #
Every punishment goes through the panel. /ban, /tempban, /mute, /tempmute, /kick, /warn, /unban, /pardon and /unmute, typed by staff or run by the console (store packages, global commands), are caught by network/Moderation/Punishments.sk whichever plugin they belong to, and sent to the panel like /nban: they apply on every server, show on the player's record and can be appealed. A time after the name is optional for bans and mutes (/ban Steve 7d Cheating; none = permanent). Lifting goes to POST /api/v1/server/punishments/revoke and is recorded with the staff member's in-game name when they have no panel account. While the panel is down, punishments and lifts apply on that server and are queued.
What players see #
What players see when banned, kicked, muted or warned (the ban screen, appeal line, ban announcement, kick screen, mute notice, mute lifted and the warning lines) is worded under Moderation → Punishment messages (permission punishments.messages), in every language, with {reason}, {duration}, {id}, {issuer}, {appeal} and {username} where they apply. They're stored as translations, so servers get them within a minute; English left empty is the scripts' own wording.
Suspicious players #
Suspicious players (Moderation → Suspicious players, /moderation/suspicious, permission suspicious.manage; App\Services\SuspiciousService, network/Moderation/Suspicious.sk): players staff are keeping an eye on, and why. Mark someone on that page (name or UUID and a reason), on their player page (Suspicious), or in-game with /suspicious <player> <reason> (alias /sus, permission network.suspicious; a linked staff account also needs suspicious.manage). /suspicious lists who's marked, online players first, and /suspicious clear <player> [note] clears a mark.
- Marking someone tells staff on every server at once (queue action
staff_alert, sent to players withnetwork.suspicious), and while they're marked, staff on a server are told whenever they join it, with the reason. - Marking someone already marked updates the reason. Clearing keeps the mark, with who cleared it and the note, under "Cleared recently" and in the audit log.
- Moderators and administrators get the permission by default.