MCP server that provides security scanning, authorization, and approval workflows for tool calls with threat detection and audit logging
Critical failures across all dimensions. Tool definitions are found only in test code (FakeMcpServer.cs), not in production server implementation. Descriptions are minimal ('Read a file.' / 'Write a file.', 12-13 chars, well below 20-char minimum). No visible input schemas, parameter descriptions, or output documentation. Tools appear to be inferred from test fixtures rather than explicitly registered in the actual MCP server. This is a non-functional evaluation baseline.
Read a file.
Write a file.
Tool definitions exist only in test code (FakeMcpServer.cs), not in production server. No explicit tool registration visible in actual MCP server implementation.
Descriptions are trivially short ('Read a file.' = 12 chars, 'Write a file.' = 13 chars). No context for LLM tool selection.
No input schemas visible. Cannot verify parameter types, constraints, or descriptions.
No output schema documentation. LLMs cannot infer what fields to expect or how to chain downstream tool calls.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 26 | 2026-07-28+ | v2 |
write_file marked as WRITE risk but no error handling, confirmation, or dry-run pattern visible. Destructive operations need recovery guidance.