Loading ToolBench
MCP server that exposes Kali Linux pentest/DAST tools over Streamable HTTP
Critical gaps across all dimensions. Tool definitions lack input schemas entirely, no JSON Schema visible in source code. Descriptions are present but minimal (10-50 chars), below the 50-200 char LLM-optimized baseline. No parameter descriptions, type definitions, or output schemas documented. Tools are security-sensitive (nmap, sqlmap, metasploit) yet lack permission gates, error recovery guidance, or audit trail documentation. No evidence of idempotency, rate limiting, or input validation patterns. Server_status is the only utility tool; others are raw command wrappers without composition support.
Web content scanner for directory enumeration
SMB/CIFS enumeration tool for Windows/Samba systems
Directory/DNS/vhost brute-force scanner
Credential brute-force tool for network services
Password cracking tool
Penetration testing framework
Web server scanner for vulnerabilities and misconfigurations
Network mapper tool for host discovery and port scanning
Check availability of installed Kali tools and server configuration
No input schemas visible in source code. All 11 tools lack JSON Schema definitions with parameter types, constraints, and descriptions. Cannot verify LLM can infer valid inputs.
Descriptions are minimal (10-50 chars, well below 50-200 char baseline). E.g., 'Network mapper tool for host discovery and port scanning' lacks WHEN to use, prerequisites, output format, or error recovery guidance. LLMs cannot reliably select or compose these tools.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | <=2025-11-25 | v2 |
SQL injection detection and exploitation tool
WordPress vulnerability scanner
No parameter descriptions or type hints visible. Tools accept raw command arguments but LLM has no guidance on format, constraints, or valid values. E.g., nmap likely accepts target, ports, scan type, none documented.
Destructive/sensitive tools (sqlmap, metasploit, hydra) lack permission gates, confirmation steps, or dry-run support. No evidence of audit logging or user authorization checks. Agents can invoke destructive operations without safeguards.
No output schema documentation. LLM cannot predict what fields each tool returns, breaking downstream tool chaining and forcing discovery detours. E.g., does nmap return JSON, CSV, or raw text?