codex-cli-docs

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). All listed URLs are official OpenAI docs or the OpenAI GitHub except for https://mcp.example.com/mcp, which is an untrusted third-party/placeholder endpoint referenced as a remote MCP and could be used to host or serve executables or resources — a potential distribution vector for malware.

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

• Potentially malicious external URL detected (high risk: 0.90). SKILL.md explicitly instructs the agent to "fetch the official page listed in the reference before answering" (e.g., https://developers.openai.com/codex/cli), which is high-confidence evidence the skill fetches remote documentation at runtime and uses that content to control prompts/answers.