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 (default MCP_URL=https://developers.openai.com/mcp) at runtime to fetch documentation pages (via search/fetch) whose text is returned and intended to be injected into the agent's context, so this external URL can directly control prompts/instructions.