mcp-registry

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The prompt embeds examples that place API tokens and DB credentials directly in config files (e.g., "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxxxx" and "postgresql://user:pass@..."), and describes adding/editing ~/.mcp.json which means an agent could be expected to read, create, or output those secret values verbatim—an exfiltration risk.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes and runs remote npm packages at runtime via npx (e.g., "npx -y @modelcontextprotocol/server-github" — https://www.npmjs.com/package/@modelcontextprotocol/server-github and "npx -y @modelcontextprotocol/server-postgres" — https://www.npmjs.com/package/@modelcontextprotocol/server-postgres), which fetches and executes external code that provides tools/resources that directly affect agent behavior.