LLM Red-Team Scanner
35-probe LLM/agent security red-team scan (injection, jailbreak, MCP abuse) with report.
我该使用它吗
质量与安全性
发现（1）
- LOWTool 'run_free_scan' description lacks action verb在 run_free_scan 中
基于对工具定义和协议合规性的自动分析。
上下文开销
这是每次将服务器的工具加载到模型上下文窗口时所消耗的大致 token 数。数值越高，可用于其他任务的注意力就越少。
安装
一键安装
将以下内容添加到你的 `claude_desktop_config.json` 文件中：
{
"mcpServers": {
"llm-red-team-scanner": {
"url": "https://llmrt-companion.manhliemcn4euwlu.workers.dev/mcp"
}
}
}远程端点
https://llmrt-companion.manhliemcn4euwlu.workers.dev/mcpstreamable-http它能做什么
工具清单
工具（4）
⚪run_free_scan(spec)
Free LLM red-team scan: 8 adversarial probes (jailbreak/prompt-injection/system-prompt-extraction/tool-abuse) against your agent or system prompt. Returns 0-100 risk score + level + full report in ~35s.
输入模式
{
"type": "object",
"properties": {
"spec": {
"type": "string",
"description": "Your agent / system prompt (>=20 chars)"
}
},
"required": [
"spec"
]
}🟢get_scan_result(job_id)
Poll a scan by job id (returned by run_free_scan) until status=done.
输入模式
{
"type": "object",
"properties": {
"job_id": {
"type": "string"
}
},
"required": [
"job_id"
]
}🟢get_pricing
PRO pricing & payment info: 24 probes / 9 attack classes / 0-100 score / CI gate, 3 USDT (ERC-20 or Lightning).
输入模式
{
"type": "object",
"properties": {}
}🟢get_purchase_flow
Machine-readable autonomous purchase procedure: how an agent pays 3 USDT and receives the PRO download with no human in the loop (callback push + poll fallback).
输入模式
{
"type": "object",
"properties": {}
}社区
证据