Security audit

网页内容获取免费版

Security checks for vulnerabilities and agentic risk

Overview

This skill is a webpage-fetching helper, but it is under-scoped and explicitly encourages using third-party services to reach crawler-filtered or Cloudflare-protected pages.

Review carefully before installing. Use it only for public pages you are authorized to fetch, and do not provide internal links, authenticated URLs, private documents, or URLs containing tokens or sensitive query parameters. The broad activation wording and Cloudflare-bypass framing make it risky for routine agent use without tighter scoping.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The skill repeatedly claims support for generic create/query/modify/delete-style operations and configurable execution modes, but the actual documented purpose is fetching webpage content through third-party services. This mismatch can cause an agent to over-trust the skill’s scope and invoke it in unintended contexts, increasing the chance of unsafe execution paths or user confusion about what data is being handled.

Vague Triggers

High

Confidence: 95% confidence
Finding: The trigger keywords include very broad terms such as API design, automation, performance optimization, conversion, writing, and marketing, which are far wider than the actual capability of fetching webpage content. This can cause the skill to activate on unrelated user requests and route sensitive or irrelevant prompts into a tool that sends URLs to external services.

Vague Triggers

High

Confidence: 94% confidence
Finding: The activation scope statement claims coverage for unrelated scenarios including marketing copy, content creation, enterprise workflows, and broad automation use cases. This ambiguity makes the skill more dangerous because agent routing may select it outside its intended narrow function, potentially exposing user-supplied URLs or causing operational misuse.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill instructs users to send target URLs to public third-party services but does not clearly warn that requested URLs and potentially sensitive query strings will be disclosed to those providers. This creates privacy and data-handling risk, especially if users provide internal, pre-release, or token-bearing links.

Ssd 2

Medium

Confidence: 88% confidence
Finding: The description frames the skill as a way to retrieve content specifically when normal crawling is filtered, which encourages use of alternative services to circumvent site-imposed access controls or anti-bot measures. Even without explicit exploit code, the documented purpose normalizes bypass-oriented behavior and can be misused against sites that intentionally block automated access.

Ssd 2

High

Confidence: 98% confidence
Finding: The Cloudflare scenario explicitly recommends using an alternate public service to access Cloudflare-protected sites, which clearly conveys bypass intent. In the skill context, this is more dangerous because the whole workflow is packaged as an agent capability with exec support, making access-control evasion easy to operationalize.

Static analysis

No suspicious patterns detected.