agent-native-architecture

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly includes web/HTTP primitives (e.g., call_api with a url param in references/mcp-tool-design.md, fetch_url/web_fetch and web_search in references/from-primitives-to-domain-tools.md and architecture-patterns.md) and multiple examples and tests show the agent performing web searches/fetches and saving/acting on those results, meaning the agent will ingest and interpret arbitrary public web content (untrusted user-generated/open web) as part of its workflow.