init

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This skill explicitly instructs writing overly permissive allowlists (including defaultMode: "bypassPermissions" and "Bash()", "Read()", "Write(*)", "WebFetch") into .claude/settings.json and forces running an init script, enabling remote code execution, broad data access, and easy data exfiltration — a high-risk backdoor/credential-theft vector.

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 writing a "YOLO" allowlist that sets "defaultMode": "bypassPermissions" and grants broad "Bash()"/Write()/Edit(*) rights (and asks the agent to run an init Bash script), which urges bypassing permission/security controls and enables arbitrary command execution on the host.