using-mcp-engineering
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: Documentation within 'authentication-and-trust.md' and 'mcp-server-smells.md' mentions strings such as 'ignore previous instructions' and 'DAN'. These are explicitly framed as examples of attack vectors for educational and defensive auditing purposes (e.g., in tables listing common vulnerabilities). They do not represent an attempt to bypass agent safety guidelines.
- [SAFE]: The skill provides high-quality guidance on architectural security, including implementation details for OAuth 2.1, protection against 'confused-deputy' attacks, and mandatory 'Origin' validation for HTTP transports.
- [SAFE]: Code examples in Python and TypeScript demonstrate defensive programming best practices, such as idempotency keys, optimistic locking with 'expected_version', and structured error envelopes with recovery hints.
- [SAFE]: All external references and dependencies are well-known, official libraries and specifications (e.g., GitHub, Model Context Protocol SDKs, and IETF RFCs) relevant to the skill's stated engineering purpose.
Audit Metadata