expert-domain-priming

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (medium risk: 0.30). The content explicitly instructs prompting the model by directly naming and impersonating real-world experts (e.g., "당신은 [실존 전문가명]입니다") to force authority-based routing signals, which constitutes suspicious steering via fabricated authority though not overtly malicious.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). SKILL.md→references/full.md의 필수 런타임 워크플로우는 “Step 2: 실제 전문가 2-3명 조회”와 “Step 5: 전문가 없을 때 폴백”에서 웹 검색/전문가 탐색(외부 텍스트)을 통해 <role>에 넣을 “찾은 전문가”의 정보(저서·프레임워크·용어)를 추출하도록 요구한다.