Smart Web Fetch Safe
PassAudited by ClawScan on May 1, 2026.
Overview
This is a coherent web-page fetching skill, with the main cautions being broad URL access and optional third-party processing through Jina in remote mode.
This skill appears safe for ordinary public web-page fetching. Keep it in local mode for privacy, configure ALLOWED_DOMAINS if you want to restrict what it can fetch, and avoid remote mode for private URLs or pages containing sensitive data.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If left unrestricted, the agent can fetch any URL it is asked or decides to fetch, including URLs the user may not have intended.
The fetcher allows any domain unless the user configures an allowlist. This is expected for a general web-fetch tool, but it is broad network authority.
if not ALLOWED_DOMAINS:
return True # 未配置白名单，允许所有Set ALLOWED_DOMAINS for sensitive environments, and avoid asking the agent to fetch private, internal, or token-bearing URLs unless you trust the workflow.
Using --remote, or setting DEFAULT_MODE=remote, can expose the target URL and fetched page content to Jina AI's reader service.
The documentation clearly discloses that remote mode sends the URL and page processing through a third-party service.
远程清洗模式：URL 和内容会经过 Jina AI 服务处理
Keep the default local mode for private or sensitive pages, and use remote mode only for public pages where third-party processing is acceptable.
Future package versions could behave differently from the versions the author tested.
The dependency installation is user-directed and uses common packages, but the documentation does not pin package versions.
pip install beautifulsoup4 requests
Install dependencies from a trusted package index and consider pinning known-good versions if using this in a controlled environment.