Loading ToolBench
This is NOT an MCP server, it is a static analysis tool (mcp-bom) that SCANS other MCP servers for security vulnerabilities. The 8 'tools' listed are test fixtures in extractor/tests/fixtures/ demonstrating dangerous patterns (filesystem access, env var leakage, secrets exposure). The actual mcp-bom package provides a CLI tool (mcp_bom.cli:main) and Python library for analyzing MCP server source code, not an MCP server itself. No tool definitions, schemas, or descriptions exist in the actual server implementation. The fixture tools are intentionally dangerous examples used for testing the detector, they are NOT production tools. Scoring reflects that this is fundamentally misclassified as an MCP server.
Delete a file from filesystem
Return service URLs — reads literal env vars for health-check endpoint.
Return any env var by name — LLM-controlled key.
Dump full process environment to the LLM — Tier 2.
List files in a directory
Return any env var by LLM-supplied name — Tier 2.
Read file contents from filesystem
This is a static analysis tool, not an MCP server. The 8 'tools' are test fixtures demonstrating dangerous patterns, not production tool definitions.
No actual MCP server implementation found. The repo contains a Python package for scanning MCP servers (mcp_bom CLI), not a server that exposes tools via MCP protocol.
Fixture tools lack proper parameter descriptions. 'path' and 'varname' parameters have minimal context; descriptions do not explain constraints, formats, or error cases.
get_connection_info and list_all_env have empty input schemas (no parameters defined), making their purpose unclear and preventing proper LLM invocation.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | <=2025-11-25 | v2 |
Write content to a file
Fixture tools expose dangerous operations (write_file, delete_file, arbitrary env var access) without permission gates, confirmation steps, or audit logging, intentional for testing but not production-safe.