Loading ToolBench
A comprehensive security auditing tool designed to detect vulnerabilities and misconfigurations in applications using the Model Context Protocol (MCP).
SecureMCP exhibits severe definition quality gaps. Tool names lack action verbs (e.g., 'scan', 'server', 'ScanAuth' mix PascalCase/lowercase inconsistently). Descriptions are present but generic, most under 100 chars, lacking WHEN/WHY context. Input schemas exist but are minimal: most tools accept only 1-2 string parameters with no validation constraints, enums, or ranges. No output schemas documented. Error handling is absent, no recovery guidance, no categorization. The 'SaveReport' tool is the only one with structured input (object type), but lacks field-level descriptions. Tool composition is poor: 'Scan' and 'ScanWithOptions' duplicate functionality; 'ScanAuth', 'ScanToken', 'ScanPrompt' should be unified under a single configurable scanner. No idempotency guarantees, no pagination, no rate-limiting guidance.
Creates a new scan report with vulnerability analysis and summary statistics
Saves the report in the specified format (JSON or HTML)
Performs a comprehensive security scan on the target MCP server including OAuth token storage, prompt injection vulnerabilities, and server authentication checks
Performs authentication security checks including SSL/TLS configuration, authentication methods, and security headers
Performs prompt injection vulnerability tests using various injection payloads
Performs security checks on an OAuth token including format validation, expiration checks, scope analysis, and storage security
Performs a security scan with specific options for OAuth, prompt injection, and authentication scanning
Inconsistent naming convention: mix of PascalCase (ScanAuth, ScanToken) and lowercase (scan, server). LLMs struggle with inconsistent casing and cannot reliably parse intent.
No output schemas documented. LLMs cannot plan downstream calls or extract structured data. Responses are untyped and unpredictable.
Duplicate/overlapping tools: 'Scan' and 'ScanWithOptions' perform the same function with different parameter styles. 'ScanAuth', 'ScanToken', 'ScanPrompt' should be unified under a single configurable scanner tool.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
Checks the security of the token endpoint including HTTPS usage and security headers
Scan an MCP server for vulnerabilities
Start the SecureMCP server
No error handling or recovery guidance. Tools lack descriptions of failure modes, retryability, or next steps. LLMs cannot self-correct on failures.
'server' tool has empty input schema and generic description. Does not convey what it does or when to call it. Violates tool-description pattern.