360 Secure SkillHub
360 安全技能中心

proxy-web-fetch

核心功能:通过OpenClaw Manager代理抓取指定URL网页内容,自动管理API密钥无需手动配置;支持转换为Markdown或纯文本,可控制缓存、图片保留、生成图文摘要,提取页面标题、关键词等元数据。典型场景:读取指定文档页、提取纯文本、强制获取最新页面等。输入为含目标URL的JSON请求(支持cURL或脚本调用),输出为结构化JSON。需配置WEB_FETCH_PROXY_URL,仅抓指定URL,不支持关键词搜索。

whyhit2005
whyhit2005
热度 66.99
最近更新 2026-05-11 17:31:38
分类:效率依赖:代码执行 / 第三方 API应用场景:文件转换 / 研究

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 proxy-web-fetch
Metadata
name
proxy-web-fetch
description
Proxy Web Page Fetch Tool - Fetches and parses web page content into structured Markdown or text via the OpenClaw Manager proxy. Use when: - Need to fetch and read the content of a web page by URL - Need to convert web pages to Markdown or plain text format - Need to extract page content with or without images - Need to get page metadata (title, description, keywords) - Need to control caching, image retention, or summary options for fetched content - User asks to "read a URL", "fetch a page", "grab the content of a web page", "scrape" or "crawl" a URL This skill routes all fetch requests through the Manager Web Fetch Proxy (configured via `WEB_FETCH_PROXY_URL` env var, required), which handles API key management automatically — no manual configuration needed. Do NOT confuse with web search — this skill fetches a specific URL's content, it does not perform keyword searches.
metadata
{"openclaw":{"requires":{"bins":["curl"],"envs":["WEB_FETCH_PROXY_URL"]}}}

Proxy Web Page Fetch

Fetch and parse web page content via the OpenClaw Manager Web Fetch Proxy. The Manager handles API key injection from encrypted storage automatically — no manual key configuration needed.

The proxy URL is configured via the WEB_FETCH_PROXY_URL environment variable (required). If not set, the skill will not be available.

Quick Start

Basic cURL Usage

curl --request POST \
  --url "${WEB_FETCH_PROXY_URL}/" \
  --header 'Content-Type: application/json' \
  --data '{
    "url": "https://www.example.com"
  }'

Script Usage

A wrapper shell script is provided for convenience.

# Basic Fetch (returns Markdown by default)
./scripts/proxy_fetch.sh --url "https://www.example.com"

# Fetch as plain text, no cache
./scripts/proxy_fetch.sh \
  --url "https://docs.python.org/3/" \
  --format text \
  --no-cache

# Fetch with image and link summaries
./scripts/proxy_fetch.sh \
  --url "https://news.example.com/article" \
  --images-summary \
  --links-summary

# Fetch without images, disable GFM
./scripts/proxy_fetch.sh \
  --url "https://blog.example.com/post" \
  --no-images \
  --no-gfm

Authentication

No authentication required — the proxy reads API keys internally from the Manager's encrypted secrets store.

API Parameter Reference

Parameter Type Required Default Description
url string ✅ - URL of the web page to fetch
timeout integer - 20 Request timeout in seconds
no_cache boolean - false Disable caching (true/false)
return_format string - markdown Return format: markdown or text
retain_images boolean - true Retain images in output (true/false)
no_gfm boolean - false Disable GitHub Flavored Markdown (true/false)
keep_img_data_url boolean - false Keep image data URLs (true/false)
with_images_summary boolean - false Include images summary (true/false)
with_links_summary boolean - false Include links summary (true/false)

Response Structure

The proxy returns JSON with the parsed page content.

{
  "id": "task-id",
  "created": 1704067200,
  "request_id": "request-id",
  "model": "model-name",
  "reader_result": {
    "title": "Page Title",
    "description": "Brief page description",
    "url": "https://www.example.com",
    "content": "Parsed page content (Markdown or text)",
    "external": {
      "stylesheet": {}
    },
    "metadata": {
      "keywords": "page, keywords",
      "viewport": "width=device-width",
      "description": "Meta description",
      "format-detection": "telephone=no"
    }
  }
}

Key Response Fields

Field Description
reader_result.content Main parsed content (body text, images, links)
reader_result.title Page title
reader_result.description Brief page description
reader_result.url Original page URL
reader_result.metadata Page metadata (keywords, viewport, etc.)

Common Use Cases

Scenario Command
Read a documentation page --url <doc_url>
Extract text only (no images) --url <url> --no-images --format text
Force fresh fetch (bypass cache) --url <url> --no-cache
Get content with all summaries --url <url> --images-summary --links-summary
Long page with extended timeout --url <url> --timeout 60

Environment Requirements

  • OpenClaw Manager must be running with the Web Fetch Proxy enabled.
  • WEB_FETCH_PROXY_URL environment variable must be set to the proxy URL (required, no default).
  • curl command must be available in your system path.