360 Secure SkillHub
360 安全技能中心

ScopeBlind protect-mcp

该 Skill 将任意 MCP 服务器包装为 stdio 代理,按工具策略对调用执行允许、拒绝、限速或人工审批,并以 Ed25519 签名收据记录决策;支持影子模式与强制模式,收据可离线校验,并附带策略包与 OWASP MCP Top 10 控制映射。

tomjwxf
tomjwxf
热度 46.43
最近更新 2026-05-11 17:37:12
分类:安全依赖:代码执行 / 读写文件 / MCP 服务应用场景:API 集成 / DevOps / 工作流 / 自动化脚本

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add scopeblind-protect-mcp
Metadata
name
scopeblind-protect-mcp
description
MCP security gateway. Wraps any MCP server with per-tool policies, Ed25519-signed decision receipts, and human approval gates. Shadow mode logs everything without blocking. Enforce mode applies policy.
metadata
{"emoji":"🛡️","requires":{"bins":["npx"],"env":[]},"install":"npm install -g protect-mcp@latest\n","license":"MIT","allowed-tools":["Bash","Read","Write"]}

protect-mcp — MCP Security Gateway

What This Skill Does

Wraps any MCP server as a transparent stdio proxy with per-tool security policies and cryptographic audit trail. Every tool call decision is logged and optionally Ed25519-signed.

Quick Start

# Shadow mode — log everything, block nothing
npx protect-mcp -- node your-server.js

# Enforce mode — apply per-tool policies
npx protect-mcp --policy policy.json --enforce -- node your-server.js

# Initialize signing (generates Ed25519 keypair)
npx protect-mcp init

Policy Example

{
  "tools": {
    "db_write": { "decision": "deny" },
    "file_read": { "decision": "allow", "rateLimit": { "maxCalls": 30, "windowSecs": 60 } },
    "deploy": { "decision": "require_approval" }
  }
}

Pre-built Policy Packs

protect-mcp ships CVE-anchored policy packs:

# List available policies
npx protect-mcp policies

# Apply the Clinejection prevention policy
npx protect-mcp --policy clinejection --enforce -- node your-server.js

Verify Receipts

Receipts are independently verifiable offline — no ScopeBlind dependency:

npx @veritasacta/verify receipt.json
npx @veritasacta/verify --self-test

OWASP MCP Top 10 Coverage

Risk Control
MCP-01 Rug Pulls Signed tool manifests; policy pins allowed tools
MCP-03 Tool Poisoning Per-tool allow/deny/rate-limit policies
MCP-04 Tool Arg Injection Argument inspection + approval gates
MCP-07 Auth/AuthZ Trust-tier gating
MCP-08 Logging & Audit Ed25519-signed receipts — verifiable offline
MCP-09 Excessive Agency Shadow mode reveals actual tool usage

Links