Deny-by-default local stdio proxy and policy inspection CLI for MCP servers with security policy enforcement, audit logging, and tool capability classification.
This is a security proxy, not a functional MCP server. The 5 tools have minimal or missing schemas and descriptions. demo_status has an empty input schema ({}), read_file and run_command lack visible parameter documentation, read_secret and unknown_tool have vague descriptions ('Do something vaguely useful'). No tool descriptions explain WHEN to use them, what they return, or error recovery paths. Schemas are not documented. This server is infrastructure for policy enforcement, not a production tool suite.
Return a fixed status message without reading files, network, environment, or arguments.
Read a file from a caller-provided path.
Read a secret reference by label.
Run a shell command.
Do something vaguely useful.
No input schemas visible for any tool. demo_status has empty object {}, others lack parameter definitions entirely. Cannot verify parameter types, constraints, or descriptions.
Descriptions are missing or trivial. 'unknown_tool' has 'Do something vaguely useful' (4 words). read_file, run_command, read_secret lack context on WHEN to use them, what they return, or error handling.
No output schemas documented. Agents cannot plan downstream calls or extract required fields. No pagination, no structured response format visible.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 35 | 2026-07-28+ | v2 |
run_command is DESTRUCTIVE but lacks confirmation/dry-run pattern. No error guidance for failed commands. Agents cannot safely retry or recover.
Tool names lack action verbs. 'unknown_tool' is not a verb_noun pattern. 'demo_status' is vague, does it return system status, demo readiness, or something else?