tool-orchestration

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow includes MCP tool discovery/calls (e.g., tools/list and tools/call) whose responses are read from an external MCP server and then routed back into the agent loop, so outsider-authored tool output can become LLM context via the tool result/pipeline propagation path.