Generated comprehensive README documentation for each cog: - dashboard (EveDash): Web dashboard setup & commands - newhelpmenu: Components V2 help system - nexuscore: All-in-one management (tickets, mod, economy, etc.) - selfbotguard: 14-layer selfbot detection - serverwhitelist: Server join management - userslash: User-installable slash commands - voicemeister: Temporary voice channels - wormhole: Cross-server relay (100+ commands) Each README includes: features, installation, commands, data statement. |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| info.json | ||
| README.md | ||
| selfbotguard.py | ||
SelfbotGuard v1.2.0
Advanced selfbot detection with 14 heuristic layers and configurable per-server punishments for Red-DiscordBot.
Detection Layers
| # | Heuristic | What it detects |
|---|---|---|
| 1 | Rich-embed abuse | Selfbots sending embeds (regular users can't) |
| 2 | Inhuman response time | Replies faster than humanly possible |
| 3 | Timing precision | Unnaturally consistent message intervals |
| 4 | 24/7 activity | No sleep/break patterns |
| 5 | Pattern matching | Automated command-response sequences |
| 6 | Burst detection | Rapid-fire message floods |
| 7 | Cross-channel spam | Same message blasted across channels |
| 8 | Cross-server activity | Messages in different servers within impossible windows |
| 9 | Edit cadence | Suspiciously regular edit timing |
| 10 | Presence anomalies | Messaging while offline/invisible |
| 11 | Reaction sniping | Reacting inhumanly fast (giveaway sniping) |
| 12 | Formatting analysis | Suspiciously clean formatting at speed |
| 13 | Trigger-word sniping | Instant response to giveaway/nitro keywords |
| 14 | Self-delete patterns | Automated self-deletion behaviour |
Each heuristic contributes to a cumulative suspicion score per user. When the score exceeds the configurable threshold, action is taken.
Requirements
- Red-DiscordBot >= 3.5.0
- Python >= 3.9
Installation
[p]repo add EveCogs https://github.com/everestmcarthur/EveCogs
[p]cog install EveCogs selfbotguard
[p]load selfbotguard
⚠️ The cog is disabled by default — enable per-server with
[p]sbguard enable.
Commands
Prefix Commands ([p]sbguard)
| Command | Description |
|---|---|
[p]sbguard enable |
Enable detection in this server |
[p]sbguard disable |
Disable detection |
[p]sbguard action <action> |
Set punishment: warn, mute, kick, ban, log |
[p]sbguard sensitivity <level> |
Set sensitivity: low, medium, high, paranoid |
[p]sbguard threshold <number> |
Set custom suspicion threshold |
[p]sbguard logchannel <#channel> |
Set the alert/log channel |
[p]sbguard muteduration <time> |
Set mute/timeout duration |
[p]sbguard exempt <@role> |
Exempt a role from detection |
[p]sbguard unexempt <@role> |
Remove exemption |
[p]sbguard notifystaff <on/off> |
Toggle staff DM notifications |
[p]sbguard settings |
View current configuration |
[p]sbguard flagged |
View flagged/suspected users |
[p]sbguard unflag <@user> |
Clear a user's suspicion score |
[p]sbguard scan <@user> |
Manually scan a specific user |
[p]sbguard reset |
Reset all detection data for this server |
Slash Commands
| Command | Description |
|---|---|
/sbguard-status |
Check SelfbotGuard status (admin only) |
/sbguard-scan <user> |
Scan a user for selfbot indicators (admin only) |
Data Statement
This cog stores Discord user IDs and guild IDs for selfbot detection. Per-user data includes message timestamps, response timing, activity windows, and suspicion scores. Data is periodically pruned.
License
MIT