openai-knowledge

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs the agent to use the openaiDeveloperDocs MCP server (https://developers.openai.com/mcp) at runtime to fetch markdown docs that the agent must base its responses on, meaning remotely fetched content from that URL would directly control prompts/instructions.