← 返回 Skills 市场
407
总下载
0
收藏
2
当前安装
1
版本数
在 OpenClaw 中安装
/install smart-web-fetch-safe
功能描述
安全版智能网页内容获取技能。本地解析 + 可选远程清洗，隐私优先，Token 优化。
使用说明 (SKILL.md)
Smart Web Fetch Safe
安全版智能网页内容获取技能，隐私优先，支持本地解析和远程清洗两种模式。
核心功能
- 本地解析默认: 使用本地 HTML 解析，隐私安全
- 可选远程清洗: 用户可选择使用 Jina Reader 远程服务
- Token 优化: 自动去除广告、导航栏等噪音内容
- 域名白名单: 可配置允许访问的域名列表
- 字符数限制: 内置最大字符数限制，避免超长输出
安全特性
⚠️ 隐私提示:
- 本地解析模式：数据完全保留在本地，隐私安全
- 远程清洗模式：URL 和内容会经过 Jina AI 服务处理
使用方式
命令行
# 本地解析模式（默认，隐私安全）
python3 skills/smart-web-fetch-safe/scripts/fetch.py "https://example.com/article"
# 远程清洗模式（更节省 Token，但数据经过第三方）
python3 skills/smart-web-fetch-safe/scripts/fetch.py "https://example.com/article" --remote
# 指定最大字符数
python3 skills/smart-web-fetch-safe/scripts/fetch.py "https://example.com/article" --max-chars 5000
# 组合使用
python3 skills/smart-web-fetch-safe/scripts/fetch.py "https://example.com/article" --remote --max-chars 3000
JSON 输出
python3 skills/smart-web-fetch-safe/scripts/fetch.py "https://example.com/article" --json
配置说明
环境变量（可选）
| 变量 | 默认值 | 说明 |
|---|---|---|
| ALLOWED_DOMAINS | * | 允许访问的域名，用逗号分隔 |
| MAX_CHARS | 10000 | 最大返回字符数 |
| DEFAULT_MODE | local | 默认模式：local（本地）或 remote（远程） |
白名单示例
export ALLOWED_DOMAINS="example.com,github.com,wikipedia.org"
export MAX_CHARS=5000
export DEFAULT_MODE=local
模式对比
| 特性 | 本地解析 (local) | 远程清洗 (remote) |
|---|---|---|
| 隐私 | ✅ 完全本地 | ⚠️ 数据经第三方 |
| Token 优化 | ✅ 基础优化 | ✅ 深度优化 50-80% |
| 速度 | 较快 | 依赖网络 |
| 依赖 | beautifulsoup4, requests | 无额外依赖 |
安装依赖
pip install beautifulsoup4 requests
当前状态
开发中。
安全使用建议
This skill appears to do what it claims: local HTML parsing by default and an optional remote cleaning call to r.jina.ai. Before using: (1) If you care about privacy, keep DEFAULT_MODE=local and install beautifulsoup4 so parsing stays local. (2) If you enable remote (--remote or DEFAULT_MODE=remote), understand the target URL will be fetched/processed by a third party (r.jina.ai) — do not send sensitive/private URLs. (3) Consider setting ALLOWED_DOMAINS to a tight whitelist (e.g., example.com, wikipedia.org) because when ALLOWED_DOMAINS is unset the script allows all domains. (4) No credentials are requested by the skill, and there is no installer that fetches arbitrary code, but review network behavior if you operate in a restricted environment.
功能分析
Type: OpenClaw Skill
Name: smart-web-fetch-safe
Version: 1.0.0
The skill bundle provides a utility for fetching and cleaning web content using either local BeautifulSoup parsing or the third-party Jina Reader service (r.jina.ai). The code in scripts/fetch.py implements standard web scraping logic with clear privacy warnings, a configurable domain whitelist to mitigate SSRF, and no evidence of malicious intent, data exfiltration, or unauthorized execution.
能力评估
Purpose & Capability
Name/description, required binary (python3), and declared deps (beautifulsoup4, requests) match the implemented behavior. The script implements local HTML cleaning and an optional remote cleaning path (Jina Reader) as described. Minor nit: SKILL.md emphasizes a "domain whitelist" but the implementation allows all domains when ALLOWED_DOMAINS is unset (documented in SKILL.md).
Instruction Scope
Runtime instructions and the script stay within the stated purpose: fetching a URL, locally cleaning HTML, or optionally calling a third-party cleaning endpoint. The only data sent to an external service in remote mode is a GET to r.jina.ai with the target URL encoded into the path (the remote service performs the fetch/clean). SKILL.md warns about third-party processing, but it says both URL and content are processed—implementation only issues a GET for the target URL (i.e., it does not POST local content).
Install Mechanism
Instruction-only skill with no install spec and a small Python script. Dependencies are standard pip packages (beautifulsoup4, requests) and no remote code is downloaded or extracted by an installer. Low installation risk.
Credentials
No required credentials or secrets are requested. SKILL.md documents optional env vars (ALLOWED_DOMAINS, MAX_CHARS, DEFAULT_MODE) which the script reads; registry shows none required—this is consistent. Caution: default ALLOWED_DOMAINS empty -> effectively allows fetching any domain unless the user sets a whitelist, which may be surprising given the emphasis on a whitelist feature.
Persistence & Privilege
always is false, no persistent installation or modifications to other skills or system config. The skill does network I/O when used but does not request elevated privileges or permanent presence.
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 smart-web-fetch-safe - 安装完成后，直接呼叫该 Skill 的名称或使用
/smart-web-fetch-safe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
Initial release of smart-web-fetch-safe:
- Provides secure web content fetching with local HTML parsing as default for privacy.
- Supports optional remote cleaning using Jina Reader for deeper content optimization.
- Implements automatic noise removal (ads, navbars), token usage optimization, and max character limits.
- Includes configurable domain whitelist for controlled access.
- Offers both command-line and JSON-output modes; flexible configuration via environment variables.
元数据
常见问题
Smart Web Fetch Safe 是什么？
安全版智能网页内容获取技能。本地解析 + 可选远程清洗，隐私优先，Token 优化。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 407 次。
如何安装 Smart Web Fetch Safe？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install smart-web-fetch-safe」即可一键安装，无需额外配置。
Smart Web Fetch Safe 是免费的吗？
是的，Smart Web Fetch Safe 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Smart Web Fetch Safe 支持哪些平台？
Smart Web Fetch Safe 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 Smart Web Fetch Safe？
由 Jory（@vimself）开发并维护，当前版本 v1.0.0。
推荐 Skills