← 返回 Skills 市场

coze-web-fetch

Name: coze-web-fetch
Author: hanxueyuan

作者 hanxueyuan · GitHub ↗ · v0.1.0 · MIT-0

cross-platform ⚠ suspicious

389

总下载

当前安装

版本数

在 OpenClaw 中安装

/install coze-web-fetch

功能描述

Fetch and extract content from URLs using coze-coding-dev-sdk. Supports web pages, PDF, Office documents (doc/docx/ppt/pptx/xls/xlsx/csv), text files, e-book...

使用说明 (SKILL.md)

Coze Web Fetch

Fetch and extract structured content from any URL using coze-coding-dev-sdk. Returns text, images, and links in various output formats.

Quick Start

Basic Fetch

npx ts-node {baseDir}/scripts/fetch.ts -u "https://example.com/article"

Multiple URLs

npx ts-node {baseDir}/scripts/fetch.ts \
  -u "https://example.com/page1" \
  -u "https://example.com/page2"

Output as Markdown

npx ts-node {baseDir}/scripts/fetch.ts \
  -u "https://docs.python.org/3/tutorial/" \
  --format markdown

Output as JSON

npx ts-node {baseDir}/scripts/fetch.ts \
  -u "https://example.com/document.pdf" \
  --format json

Text Only (No Images/Links)

npx ts-node {baseDir}/scripts/fetch.ts \
  -u "https://example.com/article" \
  --text-only

Script Options

Option	Description
`-u, --url \x3Curl>`	URL to fetch (required, can be repeated)
`--format \x3Cfmt>`	`json`, `text`, `markdown` (default: text)
`--text-only`	Extract text content only
`--help`	Show help message

Supported Document Formats

Format	Extensions
PDF	.pdf
Office Documents	.doc, .docx, .ppt, .pptx, .xls, .xlsx, .csv
Text Files	.txt, .text
E-books	.epub, .mobi
XML	.xml
Images	.jpg, .png, .gif, .webp, etc.
Web Pages	.html, .htm, or any URL

Output Formats

Text (default)

============================================================
FETCHED CONTENT
============================================================
Title: Example Article
URL: https://example.com/article

------------------------------------------------------------
CONTENT
------------------------------------------------------------
[TEXT] This is the main article content...

[IMAGE] https://example.com/image.jpg

[LINK] Related Article -> https://example.com/related

Markdown

# Example Article

**URL**: https://example.com/article

---

## Content

This is the main article content...

![Image](https://example.com/image.jpg)

- [Related Article](https://example.com/related)

JSON

Raw API response with full content structure.

Content Types

The fetcher extracts three types of content:

Type	Description
text	Extracted text content from the page
image	Image URLs with display information
link	Hyperlinks found in the content

Notes

Use --text-only for cleaner output when you only need text
PDF and Office documents are automatically parsed
Images are re-signed for secure access
Multiple URLs can be fetched in a single command

安全使用建议

This skill appears to do what it says (fetch and extract content) but there are gaps you should verify before installing: 1) The script imports 'coze-coding-dev-sdk' but the skill does not declare how/when that package or ts-node/typescript are installed — confirm the runtime will provide these dependencies. 2) The SDK likely sends fetched content to coze's API; ask whether any API key or account is required, and which environment variables or config files (if any) the SDK reads. The skill does not declare required credentials, which is a red flag for possible unintended data transmission. 3) The README claim that 'images are re-signed' is not implemented in the visible code — clarify what 're-signed' means and whether any signing keys are needed. 4) Before using on sensitive URLs, review or vendor-audit the coze-coding-dev-sdk source (or run the fetcher in an isolated environment) to see exactly what data is uploaded and where. If you need to proceed, require the author to add explicit dependency and env var declarations (and a statement of what data is sent to coze.com) or run the tool in a sandbox.

功能分析

Type: OpenClaw Skill Name: coze-web-fetch Version: 0.1.0 The skill bundle provides a legitimate utility for fetching and parsing web content and documents (PDF, Office, etc.) using the coze-coding-dev-sdk. The code in scripts/fetch.ts is straightforward, implementing command-line argument parsing and output formatting without any signs of data exfiltration, malicious execution, or obfuscation.

能力评估

⚠ Purpose & Capability

The name/description and the script align: it uses a coze SDK to fetch and extract content. However the skill requires the external package coze-coding-dev-sdk (imported in scripts/fetch.ts) but the SKILL.md and registry metadata do not declare that package or other Node dependencies (typescript/ts-node). Also the SKILL.md claims 'Images are re-signed for secure access' which is not implemented in the visible code — that claim is inconsistent.

⚠ Instruction Scope

The SKILL.md instructs running 'npx ts-node scripts/fetch.ts' which will execute the script that calls client.fetch(url). That likely sends the fetched document content to the remote coze service (via the SDK) — this network transmission is expected for this purpose but is not documented in SKILL.md (no explicit note that content is uploaded to coze.com). The instructions do not read local files or secrets, but they omit any mention of required credentials or config that the SDK's Config() may load.

ℹ Install Mechanism

There is no install spec (instruction-only), which is low risk. However required runtime packages are undeclared: the script imports 'coze-coding-dev-sdk' and is executed with ts-node; SKILL.md only lists 'npx' as required. Without explicit install steps, the script may fail or behave differently depending on the environment. This is an operational/documentation gap rather than direct evidence of malice.

⚠ Credentials

The skill declares no required environment variables or credentials, yet it instantiates Config() from an external SDK — that Config may read API keys or other tokens (e.g., COZE_API_KEY or a config file). The absence of declared env vars or primary credential is disproportionate and hides whether user data will be sent to a third-party service requiring auth. The skill also makes claims (image re-signing) that could imply use of signing keys, but no keys are declared.

✓ Persistence & Privilege

The skill is not always-enabled and does not request elevated or persistent platform privileges. It does not modify other skills or system settings in the visible files.

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 coze-web-fetch
安装完成后，直接呼叫该 Skill 的名称或使用 /coze-web-fetch 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

v0.1.0

Initial release. - Fetch and extract content from URLs using coze-coding-dev-sdk. - Supports web pages, PDF, Office documents, text files, e-books, XML, and images. - Returns structured output with text, images, and links. - Offers output in text, markdown, or JSON formats. - Allows fetching multiple URLs in a single command. - Provides options for text-only extraction.

v1.0.0

- Initial release of coze-web-fetch. - Fetch and extract structured content from web pages, PDFs, Office documents, text files, e-books, XML, and images. - Supports output as plain text, Markdown, or JSON. - Handles multiple URLs in a single command. - Option to extract text-only content. - Extracted output includes text, images, and links.

元数据

Slug coze-web-fetch

版本 0.1.0

许可证 MIT-0

累计安装 5

当前安装数 4

历史版本数 2

常见问题

coze-web-fetch 是什么？

Fetch and extract content from URLs using coze-coding-dev-sdk. Supports web pages, PDF, Office documents (doc/docx/ppt/pptx/xls/xlsx/csv), text files, e-book... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 389 次。

如何安装 coze-web-fetch？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install coze-web-fetch」即可一键安装，无需额外配置。

coze-web-fetch 是免费的吗？

是的，coze-web-fetch 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

coze-web-fetch 支持哪些平台？

coze-web-fetch 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 coze-web-fetch？

由 hanxueyuan（@hanxueyuan）开发并维护，当前版本 v0.1.0。