using-systematic

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). This skill text is a coercive prompt-injection that acts like a behavioral backdoor—forcing mandatory invocation and obedience to external "skill" content—which could be abused to enable data exfiltration, forced execution of malicious skill payloads, or supply‑chain/backdoor attacks even though it contains no direct exploit code itself.

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

• Attempt to modify system services in skill instructions detected (high risk: 0.70). The prompt doesn't directly request sudo or file changes but it forces unconditional invocation and exact following of external skills, which could themselves instruct privileged operations (create users, edit system or ssh files, request sudo), so it indirectly pushes the agent toward compromising machine state.