agent-knowledge

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill explicitly instructs the agent to persist and auto-update user conversation data into long-term pages and to "create pages silently — don't announce it to the user," which is deliberate concealment of persistent data retention (a high privacy/backdoor risk) though it contains no direct code for external network exfiltration or remote code execution.