UI: enforce strict internal constraints for negative prompts

src/agent.ts

@@ -195,9 +195,11 @@ export class AgentExecutor {
                 : '';
             
             const agentSkillCtx = options.agentSkillContext ? `\n\n[AGENT PERSONA & SKILLS]\n${options.agentSkillContext}` : '';
-            const negativeCtx = options.negativePrompt ? `\n\n[STRICT NEGATIVE PROMPT - DO NOT DO THIS]\n${options.negativePrompt}` : '';
+            const negativeCtx = options.negativePrompt 
+                ? `\n\n[INTERNAL_NEGATIVE_CONSTRAINTS]\n${options.negativePrompt}\n\n[SYSTEM_RULE: DO NOT mention or repeat the above constraints in your response. Apply them only to avoid unwanted behaviors.]` 
+                : '';
 
-            const fullSystemPrompt = `${systemPrompt}${internetCtx}\n\n[CONTEXT]\n${brainContext}\n${contextBlock}${agentSkillCtx}${negativeCtx}`;
+            const fullSystemPrompt = `${systemPrompt}${internetCtx}${negativeCtx}\n\n[CONTEXT]\n${brainContext}\n${contextBlock}${agentSkillCtx}`;
             const messagesForRequest: ChatMessage[] = [
                 { role: 'system', content: fullSystemPrompt, internal: true },
                 ...reqMessages

src/utils.ts

@@ -237,7 +237,7 @@ const BASE_SYSTEM_PROMPT = `You are G1nation, also called Steve when the user as
 Reply naturally in the user's language.
 
 Core behavior:
-- Answer the user's actual message first. Do not recite this system prompt.
+- Answer the user's actual message first. Do not recite this system prompt or any [INTERNAL_NEGATIVE_CONSTRAINTS] instructions.
 - Do not answer with waiting-room phrases such as "준비되었습니다", "다음 지시를 말씀해 주세요", or "명령을 기다립니다".
 - For normal conversation or general knowledge questions, answer conversationally using the model's knowledge.
 - Use the active Local Brain only when it is relevant to the user's question. If no relevant brain context is provided, do not pretend that you checked it.