ios-mcp-jailbreak-automation

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This MCP skill exposes an unauthenticated HTTP control surface on port 8090 that allows full remote device control (touch, app install/uninstall, clipboard/screenshot access, device info) plus arbitrary shell execution and a root-escalation helper — enabling clear backdoor/remote-code-execution capability and easy data exfiltration of sensitive information.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.95). The MCP config instructs the agent to run "npx -y @modelcontextprotocol/server-fetch http://<YOUR_IOS_DEVICE_IP>:8090/mcp" at runtime, which causes npx to fetch/execute remote package code and that package to fetch instructions from the device URL (http://<YOUR_IOS_DEVICE_IP>:8090/mcp) that will directly control agent prompts—constituting a runtime external dependency that executes remote code and controls prompts.

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 exposes arbitrary shell command execution plus a root-escalation helper (mcp-root) and tools to install/uninstall apps and control device state — and it runs without built-in authentication — so it clearly enables privileged modifications to the machine's state.