streaming-patterns

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

• Third-party content exposure detected (high risk: 0.80). The skill includes streaming tools that accept arbitrary external sources (e.g., monitor_server_health(server_url) and search_documents(...) referencing "external_api" in templates/streaming-tool-template.py and examples/streaming-tool-agent.py) and yields their results into the agent event stream (event.tool_call.result), so the agent will read/interpret untrusted third-party content (URLs/external API results) at runtime.