Security overview

Phantom's security module is 13 separate features grouped under one dashboard menu. Each one targets a specific attack class — raids, nuke attempts, scam DMs, fresh-account joins, role-permission grants, etc.

The 13 features

Feature	Defends against
Lockdown	One-click "freeze chat" for incident response
Anti-Raid	Mass-join attacks
Anti-Nuke	A compromised admin / bot deleting your server
Anti-Scam	Phishing / crypto-scam image posts (vision-AI)
Verification	Bot accounts and casual raiders
Firewall	Composable join-time filters (age, avatar, bot account, etc.)
Quarantine	Strip roles + isolate a member without banning
Server Scanner	Find misconfiguration before an attacker does
Server Backup	Snapshot roles/channels/emoji for restoration (not yet generally available)
Audit Log	Tamper-resistant event trail of every security action
Honeypot	Hidden channels that auto-action crawlers
Perm Watch	Alert on integration added, webhook created, etc.
Blacklist	Auto-action specific user IDs on join

Safeguarding sits outside this list

Safeguarding is Phantom's child-safety layer: the always-available /psafety reporting commands, the refusal layer around Phantom's AI features, join-time server vetting, and the platform review queue. It is not a per-server feature and server admins cannot configure or disable any of it, which is why it isn't in the table above.

All ship disabled by default

Every security feature is off out of the box. You enable the ones that match your threat model. A small community might only ever turn on Verification + Anti-Raid; a high-profile community might use the whole suite.

Where it lives

Dashboard: Security → <feature>. Each feature has its own page.

Permissions

• security.view — see the page + current state
• security.edit — toggle / configure / take security-side actions (e.g. release a quarantined member, end a raid session)

A user with only security.view sees the dashboard pages as read-only — the save buttons are disabled with a tooltip.

Cross-Server Sync

Three of the security features have syncable settings:

• Anti-Raid (thresholds, action, account-age gate)
• Anti-Scam (sensitivity, timeout, DM-on-detect)
• Anti-Nuke (rate limits, punishment, auto-revert)

Plus the entire firewall rule list can sync per-rule.

Per-server channel + role choices stay local (each server has its own #mod-alerts and its own @Moderator role). See Cross-Server Sync for details.

Recommended starting setup

For an average mid-sized community:

1. Anti-Raid ON at default thresholds (8 joins in 10s → quarantine).
2. Anti-Scam ON at sensitivity = medium.
3. Anti-Nuke ON, action = strip_roles, auto_revert = on.
4. Verification ON, mode = button, deploy the panel in #welcome.
5. Audit Log ON, channel = #security-log.

That's the 80/20 of security. Add the rest as you discover threats specific to your community.

Recovery features

Three features focus on after-the-fact recovery:

• Server Backup — snapshot before risky changes.
• Quarantine — temporary isolation that's faster + reversible vs banning.
• Audit Log — when something went wrong, this is where you reconstruct what happened.

Keep at least one fresh backup per week; Anti-Nuke can auto-revert role changes but cannot recover from "the admin deleted the wrong channel" without a backup.

Limits (across the suite)

• 25 server backups per guild (older ones auto-prune)
• 5,000 quarantined users tracked at once
• 5,000 blacklist entries per guild
• 50 firewall rules
• 365-day audit-log retention
• 8 MB max payload for one server backup

What about automod?

Automod (under Moderation) is the message-filter side of defence. Security focuses on events (joins, role grants, channel deletes, etc.). The two complement each other; high-security servers use both.