openai-knowledge

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime path uses an MCP server to fetch OpenAI developer documentation markdown (mcp__openaiDeveloperDocs__fetch_openai_doc), which is outsider-authored public web content ingested as readable text into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 1.00). This skill requires using the MCP tools to fetch OpenAI docs at runtime (e.g., via the server URL https://developers.openai.com/mcp), and those fetched markdown docs are injected into the agent context and directly control the agent's responses.