Document the raid protection, anti-nuke, and spam protection systems Adds a Safety Systems page covering all three: what each detects, how they respond, the bot permissions they need, and guidance on sizing the thresholds against a server's own join history rather than a guess. Records the reasoning behind the non-obvious choices, since they look like oversights otherwise: link spam is detected behaviorally instead of by domain blocklist (scam domains burn within days), avatar clustering hashes image bytes rather than Discord's avatar key (whether identical uploads share a key is undocumented and unverified), and the raid lockdown uses Discord's self-expiring incident actions so no local state has to survive a restart. Spells out that anti-nuke deliberately does NOT exempt moderators. A compromised moderator account is the threat it exists for, so exempting trusted accounts would defeat it — and a reader assuming the usual moderator exemption would draw exactly the wrong conclusion. Home, Setup-Guide and Commands-Reference updated to match, including the three permissions an existing install won't already have granted.