Security audit

Web Content Fetcher

Security checks across malware telemetry and agentic risk

Overview

This skill fetches webpages through clearly named third-party relay services, but users should avoid using it with private or secret-bearing URLs.

Install only if you are comfortable with requested webpage URLs and returned page content being processed by the named external services. Do not use it for private dashboards, intranet pages, signed links, password-reset links, authenticated pages, or URLs containing tokens or other secrets.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Vague Triggers

Medium

Confidence: 78% confidence
Finding: The trigger phrases are broad enough to match common browsing and summarization requests, which can cause the skill to activate in situations where the user did not intend to use third-party relay services. In this skill's context, that matters because activation can send requested URLs to external fetch providers, expanding data exposure beyond normal browsing behavior.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill instructs the agent to send user-supplied URLs to third-party services (`r.jina.ai`, `markdown.new`, `defuddle.md`) without a clear warning or consent step. This is dangerous because URLs may contain sensitive query parameters, internal endpoints, auth tokens, or private resources, and the skill is explicitly positioned as a way to bypass filtering/Cloudflare, which increases the chance of misuse or policy circumvention.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The script sends the user-provided URL to an external proxy service (r.jina.ai) without any explicit disclosure, consent, or warning that the requested target will be shared with a third party. In this skill’s context, that behavior is the core feature, but it still creates a real privacy and data-governance risk because users may pass sensitive internal, authenticated, or confidential URLs assuming the tool fetches them directly.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: In multi-fetch mode, the same user-supplied URL is transmitted to multiple third-party services, increasing exposure beyond a single provider. This compounds privacy risk and potential leakage of sensitive URLs or query parameters, especially because the skill is explicitly designed to bypass normal filtering and may be used on protected or internal resources.

Missing User Warnings

Medium

Confidence: 98% confidence
Finding: The tool is explicitly designed to send user-supplied URLs to third-party services such as r.jina.ai, markdown.new, and defuddle.md, which exposes target URLs and related request metadata to external operators. In a security-sensitive agent skill, this can leak internal URLs, private resources, tokens embedded in query strings, or confidential browsing targets without clear user consent.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.