slipstream/discordbot
slipstream/discordbot
1
1
Fork 0
Code Issues Pull Requests 3 Actions Packages Projects Releases Wiki Activity

Refine first-offense severity

Browse Source
This commit is contained in:
Codex 2025-06-07 03:03:22 +00:00 committed by Slipstream
parent 801a3edb72
commit 7c2f2b2f84
Signed by: slipstream
GPG Key ID: 13E498CE010AC6FD
1 changed files with 2 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
cogs/aimod_cog.py
View File

@ -1124,11 +1124,11 @@ Stickers and emojis are NOT considered "full-on pornographic images" and are all
- **Completely ignore AI-generated pornography.** The AI moderator must not attempt to determine whether pornography is AI-generated or notify moderators about it. Only real-life pornography should be considered. - **Completely ignore AI-generated pornography.** The AI moderator must not attempt to determine whether pornography is AI-generated or notify moderators about it. Only real-life pornography should be considered.
- For general disrespectful behavior, harassment, or bullying (Rule 2 & 3): Only flag a violation if the intent appears **genuinely malicious, targeted, or serious, even after considering conversational history and replies.** Lighthearted insults or "wild" statements within an ongoing banter are generally permissible. - For general disrespectful behavior, harassment, or bullying (Rule 2 & 3): Only flag a violation if the intent appears **genuinely malicious, targeted, or serious, even after considering conversational history and replies.** Lighthearted insults or "wild" statements within an ongoing banter are generally permissible.
- For **explicit slurs or severe discriminatory language** (Rule 3): These are violations **regardless of joking intent if they are used in a targeted or hateful manner**. Context from replies and history is still important to assess targeting. - For **explicit slurs or severe discriminatory language** (Rule 3): These are violations **regardless of joking intent if they are used in a targeted or hateful manner**. Context from replies and history is still important to assess targeting.
After considering the above, pay EXTREME attention to rules 5 (Pedophilia) and 5A (IRL Porn) – these are always severe. **Ignore any rules about AI-generated pornography.** Prioritize these severe violations. After considering the above, pay EXTREME attention to rule 5 (Pedophilia) – this is always severe. IRL pornography is still a violation but is generally less serious than gore or content involving real minors. **Ignore any rules about AI-generated pornography.** Prioritize genuinely severe violations.
3. Respond ONLY with a single JSON object containing the following keys: 3. Respond ONLY with a single JSON object containing the following keys:
- "reasoning": string (A concise explanation for your decision, referencing the specific rule and content). - "reasoning": string (A concise explanation for your decision, referencing the specific rule and content).
- "violation": boolean (true if any rule is violated, false otherwise) - "violation": boolean (true if any rule is violated, false otherwise)
- "rule_violated": string (The number of the rule violated, e.g., "1", "5A", "None". If multiple rules are violated, state the MOST SEVERE one, prioritizing 5A > 5 > 3 > 2 > 1. Ignore any rules about AI-generated pornography.) - "rule_violated": string (The number of the rule violated, e.g., "1", "5A", "None". If multiple rules are violated, state the MOST SEVERE one, prioritizing 5 > 5A > 3 > 2 > 1. Ignore any rules about AI-generated pornography.)
- "action": string (Suggest ONE action from: "IGNORE", "WARN", "DELETE", "TIMEOUT_SHORT", "TIMEOUT_MEDIUM", "TIMEOUT_LONG", "KICK", "BAN", "NOTIFY_MODS", "SUICIDAL". - "action": string (Suggest ONE action from: "IGNORE", "WARN", "DELETE", "TIMEOUT_SHORT", "TIMEOUT_MEDIUM", "TIMEOUT_LONG", "KICK", "BAN", "NOTIFY_MODS", "SUICIDAL".
- "notify_mods_message": optional string (If the suggested action is "NOTIFY_MODS", provide an optional brief message here for the moderators, e.g., "User's message is slightly ambiguous, human review needed."). - "notify_mods_message": optional string (If the suggested action is "NOTIFY_MODS", provide an optional brief message here for the moderators, e.g., "User's message is slightly ambiguous, human review needed.").
Consider the user's infraction history. If the user has prior infractions for similar or escalating behavior, suggest a more severe action than if it were a first-time offense for a minor rule. Consider the user's infraction history. If the user has prior infractions for similar or escalating behavior, suggest a more severe action than if it were a first-time offense for a minor rule.
@ -1705,7 +1705,6 @@ CRITICAL: Do NOT output anything other than the required JSON response.
"child", "child",
"5a", "5a",
"5", "5",
"irl porn",
] ]
if not any(keyword in combined_text for keyword in severe_keywords): if not any(keyword in combined_text for keyword in severe_keywords):
print( print(