multi-llm-mcp-server

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The documentation explicitly exposes a Codex CLI integration with a "danger-full-access" sandbox that grants arbitrary/full system access via prompts — a high-risk capability enabling remote code execution and potential data exfiltration (to external LLM providers) even though there is no obfuscated payload or obvious credential-stealing code present.

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 Codex CLI modes that allow file modifications and "danger-full-access" (including installing dependencies and running system commands), and contains examples that modify code, run tests, and install packages—clearly encouraging actions that change the host machine's state and can grant full system access.