langchain-mcp-adapters

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

• Third-party content exposure detected (high risk: 0.90). The skill's SKILL.md shows connecting to remote MCP servers (e.g., StreamableHttpConnection with "url": "https://my-mcp-server.example.com/mcp") and explicitly calls load_mcp_prompt, load_mcp_resources/get_mcp_resource and get_tools to ingest prompts, resources and tools from those servers which the agent will read and act on, exposing it to untrusted third‑party content that could inject instructions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes runtime StreamableHttp/SSE connection examples that fetch prompts/resources from external MCP servers (e.g., https://weather-mcp.example.com/mcp and https://my-mcp-server.example.com/mcp / https://my-mcp-server.example.com/sse), and those fetched resources/prompts are loaded at runtime and directly control agent prompts/instructions.