openai-docs-skill

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's CLI calls the MCP endpoint (https://developers.openai.com/mcp) at runtime (via curl/json-rpc) to fetch doc text that is then output/injected for use in responses, so remote content can directly control agent instructions.