An LLM-powered penetration testing tool that provides MCP-based penetration testing tools including port scanning and web content discovery
This server has significant quality gaps across all dimensions. Only 2 tools are exposed, both with minimal schema documentation and descriptions in Korean that are too vague for LLM reasoning. Input parameters lack type definitions in the visible schema (only string types noted, no enums or constraints). No output schemas are documented. Error handling is absent. The tools themselves are network security scanners (nmap, dirb) exposed directly without safeguards, validation, or recovery guidance. Naming follows verb_noun convention (nmap_scan, dirb_scan) which is correct, but descriptions are non-English and lack the specificity needed to guide LLM tool selection. No evidence of parameter validation, permission gating, or actionable error messages.
Dirb를 사용하여 웹 서버의 디렉터리 및 파일을 열거합니다.
Nmap을 사용하여 지정된 타겟에 대한 포트 스캔을 수행합니다.
Descriptions are in Korean and lack actionable context for LLM selection. No English translations provided. Descriptions do not explain WHEN to use each tool vs. the other, what they return, or prerequisites.
No output schemas documented. LLMs cannot infer what fields nmap_scan or dirb_scan return, breaking downstream tool chaining and forcing manual result parsing.
Input parameters lack type constraints (no enums, ranges, or patterns). 'targets' and 'target_url' are bare strings with no validation hints. 'options' accepts arbitrary strings, inviting malformed nmap/dirb arguments.
No error handling or recovery guidance. If nmap or dirb fails (invalid target, command not installed, permission denied), LLMs receive raw errors with no actionable guidance on what to do next.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 35 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 37 | - | v1 |
Security concern: no permission gating or authorization checks visible. Network scanning tools (nmap, dirb) are potentially destructive and should require explicit authorization. No audit trail, rate limiting, or scope declarations.
Parameter descriptions are vague and include example values (e.g., '10.0.2.5', 'http://10.0.2.5') which LLMs may reuse literally. Should use format constraints instead.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite nmap_scan and dirb_scan being read-only tools. While the provided Risk field shows READ_ONLY, this is not part of the MCP protocol and may not be respected by clients.