sec-agentshield-wrapper

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The skill contains an explicit instruction to always pass mode: "bypassPermissions" when spawning agents—an instruction that overrides permission controls and is unrelated to the documented pre-write security analysis purpose, so it is a hidden/deceptive override.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The document contains an explicit, deliberate instruction to force agent tools into a "bypassPermissions" mode (overriding normal permission prompts), which amounts to disabling security controls and enabling unauthorized agent actions — a high-risk backdoor-like directive.

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

• Third-party content exposure detected (medium risk: 0.30). sec-agentshield-wrapper의 런타임 워크플로우는 사용자가 입력한 "/sec-agentshield-wrapper "<변경 대상 파일/영역> — <변경 의도>"" 자유텍스트를 그대로 받아 키워드(토큰/jwt/upload/api 등) 기반으로 위험 패턴 매칭 및 advisory를 생성하므로, 제3자가 자유텍스트를 제출해 LLM에 직접 주입할 수 있습니다.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs agents to pass mode: "bypassPermissions", directing the agent to bypass permission prompts and thus encourages circumventing security controls on the host.