cyber-risk-modeling

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt contains explicit self-evolution telemetry instructions—instructing the agent to search for a local project memory directory and append execution metadata (and to "skip silently" if not found)—which is a stealthy, out-of-scope data-logging behavior unrelated to the declared cyber-risk-modeling purpose.