Intent classifier for MCP servers. Goes beyond pattern matching: parses tool registrations, traces data/exec/network flow, and emits a benign/suspicious/malicious verdict with OWASP MCP Top 10 mapping.
This is a security scanner tool, not a production MCP server. The three tools (add, multiply, divide) are minimal calculator functions used as test fixtures. All three lack parameter descriptions (null values visible in schema), violating the critical requirement that every parameter must have a description. Tool descriptions are present but extremely brief (10-34 chars). No output schemas are documented. The tools are READ_ONLY and trivial, but the definition quality is poor by production standards. This server appears designed to analyze OTHER MCP servers for security issues, not to provide useful agent capabilities.
Return the sum of two integers.
Return a / b. Returns 0 if b is zero.
Return the product of two integers.
All parameters lack descriptions. Schema shows null for parameter descriptions on all three tools (a, b parameters).
No output schemas documented. Tools return int or float but LLMs cannot infer return structure. Rubric requires: 'Document the output schema. LLMs need to know what fields to expect.'
Tool descriptions are extremely brief (10-34 chars for add/multiply, 50 for divide). Rubric baseline: average tool description is 194 chars (p10=34, p90=392). These descriptions lack context for LLM selection and do not explain WHEN to use each tool.
No error handling guidance. divide() returns 0 on division by zero, but this behavior is not documented in the tool description. LLMs cannot infer this edge case handling.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | 2026-07-28+ | v2 |
Parameter naming lacks type suffixes. Parameters 'a' and 'b' are generic and do not indicate they are numeric operands. Rubric: 'When a parameter could be an ID, name, email, or position, suffix it with the type.' Use 'operand_a' or 'left_operand' for clarity.