agentic-infra

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

• Potentially malicious external URL detected (high risk: 0.90). The content shows agents fetching discovery and tool/prompt definitions at runtime (A2A Agent Card discovery at /.well-known/agent.json and MCP over HTTP/SSE), e.g. the example agent endpoint https://api.myagent.com/.well-known/agent.json (and https://api.myagent.com/a2a/tasks), which would deliver remote agent/tool/prompt definitions that directly control prompts/behavior—this is a runtime external dependency that can control agent instructions.