dex-add-mcp

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md "Examples" explicitly shows adding remote HTTP MCP endpoints (e.g., "claude mcp add --transport http https://mcp.notion.com/mcp" and "https://api.githubcopilot.com/mcp/"), which lets the agent connect to arbitrary public URLs and ingest runtime content from third‑party servers that could include instructions and enable indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes examples that add remote HTTP MCP servers (e.g., https://mcp.notion.com/mcp and https://api.githubcopilot.com/mcp/) which are invoked at runtime as MCP endpoints and therefore can deliver instructions or execute logic that directly control the agent.