BypassGuard — quarantine, not punishment
Most anti-raid bots (Wick, Dyno, and others) act as an executioner: mute, kick, ban, automatically, the moment a rule fires. That works great for raids, but strict regex filters on Unicode/homoglyph tricks throw real false positives at real users. BypassGuard takes a different approach: human-in-the-loop quarantine, zero automated bans.
How it actually catches evasion, not just banned words:
Traditional auto-mod runs regex directly on the raw message. Spammers walk right past it with BIDI override characters, zero-width spaces, or mixed Cyrillic/Greek homoglyphs. BypassGuard sanitizes and normalizes the text BEFORE evaluating it — so “h4ck3r”, stretched-out letters, reversed text, and lookalike-alphabet tricks all resolve to the same normalized form a rule can actually catch.
- Leetspeak, letter-stretching, phonetic matches, acrostics, ASCII-art patterns
- A borderline/1-letter-variant term gets quarantined for staff review, not auto-actioned
- Every ban needs a human moderator’s confirmation, one click, always
- Anti-scam honeypot trap (Pro+) — content-agnostic, catches image-only scams too
- Spanish, English, and Portuguese, built in
Not a replacement for your anti-nuke bot — a specialized anti-evasion layer for the messages that slip past native AutoMod and traditional filters.
Free tier available, no commitment.
