autobrowse-cc

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The run script constructs and sends full task prompts (including example passwords) to an external LLM process (claude -p) while enabling Bash tool access and permissive edit/add-dir options, which together allow remote command execution and easy exfiltration of secrets — a high-risk abuse/backdoor capability.