← 返回 Skills 市场
137
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install scopeblind-protect-mcp
功能描述
MCP security gateway. Wraps any MCP server with per-tool policies, Ed25519-signed decision receipts, and human approval gates. Shadow mode logs everything wi...
使用说明 (SKILL.md)
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
安全使用建议
This skill appears internally consistent: it’s a CLI tool distributed via npm and the SKILL.md uses npx/npm in ways that match the description. However, installing and running third-party npm packages executes remote code and the tool will generate signing keys and write receipts to disk. Before installing:
- Verify the npm package page, maintainers, and recent release history (don’t blindly run @latest).
- Inspect the package source (npm view, GitHub repo) or run it in an isolated container or VM first.
- Prefer a specific pinned version rather than @latest to avoid supply-chain surprises.
- Find out where the tool stores private keys/receipts and protect those files (restrict filesystem permissions, back them up if needed).
- Consider running 'npx protect-mcp --help' and reviewing docs on scopeblind.com and the npm listing before global install.
If you cannot verify the package authorship or are uncomfortable granting filesystem execution rights, run the tool in an isolated environment or decline installation.
能力评估
Purpose & Capability
The name/description match the runtime instructions: the SKILL.md shows how to run 'protect-mcp' via npx/npm, configure policies, and initialize signing. Declaring npx in metadata and using an npm install is proportionate to a CLI tool delivered via the npm ecosystem.
Instruction Scope
Instructions tell the agent to run npx protect-mcp commands that wrap a server process, generate Ed25519 keypairs, and write receipts; they do not ask for unrelated system files or credentials. However the SKILL.md does not specify where keys/receipts are stored or how long they persist, which is important because secrecy and local file writes are involved.
Install Mechanism
The install recommendation uses npm (npm install -g protect-mcp@latest) and runtime uses npx. This is a standard distribution mechanism for CLI tools but it does mean remotely-published code will be downloaded and executed on the host. No install tarball from a random URL is used, but npm packages can still contain arbitrary install scripts — verify the package source/maintainers before installing globally.
Credentials
The skill requests no environment variables or external credentials, which aligns with its intended local-proxy functionality. Caveat: the tool will generate and persist cryptographic keys and receipts locally (not declared as env/config paths), so consider where those secrets land and who can read them.
Persistence & Privilege
always:false (normal). The allowed-tools include Bash/Read/Write, which are necessary for running the proxy and creating keys/files. Installing the package globally (npm -g) writes to the system/npm directories and may require elevated permissions on some systems — run with care or use a container/virtualenv.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install scopeblind-protect-mcp - 安装完成后，直接呼叫该 Skill 的名称或使用
/scopeblind-protect-mcp触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
scopeblind-protect-mcp 1.0.0 — Initial Release
- Wraps any MCP server as a security gateway with per-tool policies.
- Supports both shadow mode (logs all actions, does not block) and enforce mode (applies policy).
- Provides cryptographically signed (Ed25519) decision receipts for all tool calls.
- Includes human approval gates, rate-limits, and per-tool allow/deny decisioning.
- Integrates pre-built, CVE-anchored policy packs.
- Receipts are verifiable offline with no external dependency.
元数据
常见问题
ScopeBlind protect-mcp 是什么？
MCP security gateway. Wraps any MCP server with per-tool policies, Ed25519-signed decision receipts, and human approval gates. Shadow mode logs everything wi... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 137 次。
如何安装 ScopeBlind protect-mcp？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install scopeblind-protect-mcp」即可一键安装，无需额外配置。
ScopeBlind protect-mcp 是免费的吗？
是的，ScopeBlind protect-mcp 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
ScopeBlind protect-mcp 支持哪些平台？
ScopeBlind protect-mcp 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 ScopeBlind protect-mcp？
由 TJF（@tomjwxf）开发并维护，当前版本 v1.0.0。
推荐 Skills