首页 › 龙虾技能列表 › Proxy4agent

Proxy4agent

v1.6.3

Residential proxy for AI agents — fetch any URL through 2M+ real IPs, bypass anti-bot, geo-target, sticky sessions.

0· 55·0 当前·0 累计
by @goldentrii (Memijashi)·MIT-0
下载技能包
License
MIT-0
最后更新
2026/4/10
安全扫描
VirusTotal
无害
查看报告
OpenClaw
可疑
medium confidence
The skill's stated purpose (residential proxy for agents) matches most of its instructions, but there are several inconsistencies and privacy/exfiltration risks (unverified npm package install, mismatched/undocumented credentials, and routing arbitrary requests through a third-party) that warrant caution.
评估建议
This skill appears to be a normal residential-proxy integration, but exercise caution before installing. Verify the npm package source (bestproxy4agents) and the GitHub repo/release tarball to ensure you trust the publisher. Clarify the credential requirements (NOVADA_PROXY_USER vs NOVADA_PROXY_PASS and optional keys) and the exact package version (1.6.2 vs 1.6.3). Remember that the proxy provider can see full request/response data — do not send internal URLs, API keys, or PII through it. If you...
详细分析 ▾
ℹ 用途与能力
The name/description align with the runtime instructions (installing an npm proxy client and providing proxy credentials). However there are small mismatches: registry metadata shows version 1.6.3 while SKILL.md references 1.6.2, and primaryEnv lists only NOVADA_PROXY_USER yet the example requires NOVADA_PROXY_PASS as well. These inconsistencies are unexplained but could be benign (out-of-date docs).
⚠ 指令范围
SKILL.md instructs the agent to install and run an npm package (bestproxy4agents) and then fetch arbitrary URLs through a third-party residential network. That is within the claimed proxy purpose, but it also means any URL (including internal or sensitive endpoints if mistakenly used) and request/response content will transit a provider you do not control — a high-risk data-exposure capability. The doc warns about this, but the agent will have the technical ability to route any data through the provider.
ℹ 安装机制
Installation uses npx to pull an npm package (bestproxy4agents). Pulling code from the public npm registry is common for this functionality but is a moderate risk because arbitrary remote code will be fetched and executed at install/runtime. The registry's declared install (node formula bestproxy4agents@1.6.2) and the SKILL.md npx usage are consistent with an npm-based install, but the version mismatch noted above should be clarified.
⚠ 凭证需求
The skill declares a primary credential NOVADA_PROXY_USER but the examples and optional features reference additional environment variables (NOVADA_PROXY_PASS, NOVADA_API_KEY, NOVADA_BROWSER_WS, PROXY_URL). The SKILL.md expects a password or proxy URL in practice; the registry metadata does not list these as required. Requesting provider credentials is expected for a proxy skill, but the metadata/instructions are inconsistent about which variables are required and which are optional.
✓ 持久化与权限
always:false and no required config paths are declared. The skill does not request elevated platform privileges or permanent forced inclusion. Autonomous invocation is allowed (the platform default), which increases the blast radius if the skill is misused, but this alone is not an incoherence.
安全有层次，运行前请审查代码。

License

MIT-0

可自由使用、修改和再分发，无需署名。

运行时依赖

🖥️ OSmacOS · Linux · Windows

版本

latestv1.6.32026/4/10

- Updated to use the latest npm package: bestproxy4agents@1.6.2. - Simplified setup instructions to require credentials for only one provider (Novada recommended). - Removed unnecessary provider environment variables from setup (BrightData, Smartproxy, Oxylabs now use PROXY_URL format). - Clarified that extra credentials (NOVADA_API_KEY, NOVADA_BROWSER_WS) are optional for search/render features. - Added a clear privacy note about proxy provider data visibility. - Improved tool documentation and updated provider options for clarity.

● 无害

安装命令 点击复制

官方npx clawhub@latest install proxy4agent
镜像加速npx clawhub@latest install proxy4agent --registry https://cn.clawhub-mirror.com

技能文档

Residential proxy MCP server for AI agents. Route HTTP requests through 2M+ real home devices to bypass anti-bot systems, geo-target by country or city, and maintain sticky sessions.

npm package: bestproxy4agents | GitHub: Goldentrii/proxy4agent

Setup

You only need credentials for ONE provider. Novada is the default and recommended provider.

Option 1: Novada (recommended)

claude mcp add proxy4agent \
  -e NOVADA_PROXY_USER=your_username \
  -e NOVADA_PROXY_PASS=your_password \
  -- npx -y bestproxy4agents

Get credentials: novada.com → Dashboard → Residential Proxies (30 seconds, no credit card).

Option 2: Any HTTP proxy

claude mcp add proxy4agent \
  -e PROXY_URL="http://user:pass@host:port" \
  -- npx -y bestproxy4agents

Works with BrightData, Smartproxy, Oxylabs, IPRoyal, or any standard HTTP proxy. Encode targeting in the URL per your provider's format.

Optional additional credentials

These are NOT required. Only set them if you need the specific feature:

  • NOVADA_API_KEY — enables agentproxy_search (Google search via Novada Scraper API)
  • NOVADA_BROWSER_WS — enables agentproxy_render (JS rendering via Novada Browser API)

Privacy note

All HTTP requests are routed through your chosen proxy provider's residential network. Do not send sensitive internal URLs, API keys, or personally identifiable information through the proxy. The proxy provider can see request URLs and response content in transit. Use only for public web content.

Tools

agentproxy_fetch

Fetch any URL through a residential proxy. Works on Amazon, LinkedIn, Cloudflare-protected sites.

agentproxy_fetch(url="https://example.com", country="US", city="newyork", format="markdown")

Parameters: url (required), country, city, session_id, format (markdown/raw), timeout (1-120s).

agentproxy_session

Sticky session — same IP across multiple requests. Use for login flows, paginated scraping.

agentproxy_session(session_id="job001", url="https://example.com/page/1")
agentproxy_session(session_id="job001", url="https://example.com/page/2")

agentproxy_search

Structured Google search. Returns titles, URLs, descriptions as clean JSON.

agentproxy_search(query="residential proxy for AI", num=5, country="us")

Requires: NOVADA_API_KEY (optional — only if you need search)

agentproxy_render

Render JS-heavy pages with real Chromium (SPAs, React/Vue apps).

agentproxy_render(url="https://react.dev", wait_for=".main-content", format="markdown")

Requires: NOVADA_BROWSER_WS (optional — only if you need JS rendering)

agentproxy_status

Check proxy network health. No credentials needed.

When to use proxy4agent

  • Target site blocks your agent (403, captcha, Cloudflare challenge)
  • Need content from a specific country or city
  • Multi-step workflow requires same IP (login + scrape)
  • JS-heavy page returns blank without a real browser
  • Need structured search results without HTML parsing
数据来源：ClawHub ↗ · 中文优化：龙虾技能库
OpenClaw 技能定制 / 插件定制 / 私有工作流定制

免费技能或插件可能存在安全风险，如需更匹配、更安全的方案，建议联系付费定制

了解定制服务