Security audit

web-crawler

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed web-crawling skill that fetches user-chosen pages and saves crawl results locally, with no hidden exfiltration or destructive behavior found.

Install only if you want a local web crawler that can contact external websites, render pages in Chromium, optionally execute user-supplied JavaScript in a page, and save crawled content/checkpoints on disk. Use same-domain mode unless you intentionally need broader crawling, set an output directory deliberately, and avoid running custom JavaScript on sensitive or authenticated pages.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (8)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 70% confidence
Finding: Without declared permissions the skill's intent is opaque and cannot be validated.

Description-Behavior Mismatch

Medium

Confidence: 91% confidence
Finding: The manifest describes whole-site BFS crawling, which implies staying within the target site. The `--no-same-domain` option explicitly allows the crawler to leave the original domain and traverse other domains, expanding behavior beyond the described site-bounded scope.

Context-Inappropriate Capability

Medium

Confidence: 85% confidence
Finding: The manifest describes page fetching/crawling with selector extraction and rendering, but this code also exposes a general-purpose `js` parameter for arbitrary script execution in the loaded page. While some rendering support is expected, allowing freeform code execution inside the browser goes beyond a normal fetch/crawl interface and is not clearly justified by the stated purpose.

Missing User Warnings

Low

Confidence: 83% confidence
Finding: This markdown file documents that the crawler automatically saves checkpoint files and stores crawl results, but it does not provide an explicit warning about local file creation or persistence of collected content. For a markdown skill description, file-writing behavior that may retain user-collected data should be disclosed clearly so users understand the side effect before running the skill.

Natural-Language Policy Violations

Medium

Confidence: 91% confidence
Finding: The skill metadata description forces a specific language presentation without any opt-in or alternative language option. Under the language/locale policy, natural-language skill content should not impose a language choice unless the constraint is explicitly justified or user-selectable.

Missing User Warnings

Low

Confidence: 82% confidence
Finding: This code launches a browser and fetches the provided URL, which is a network operation that can contact external systems and transmit standard browser metadata such as user agent and IP. While the file has terse docstrings and progress prints, it does not include a clear user disclosure warning near the operation about making outbound requests to remote sites.

Missing User Warnings

Low

Confidence: 89% confidence
Finding: The skill creates an output directory and later saves checkpoints and extracted page content as JSON, which affects user data and local storage. Although there are save-related print statements, the code does not clearly warn upfront that crawled content and resume state will be stored on disk by default.

Natural-Language Policy Violations

Medium

Confidence: 95% confidence
Finding: The code hard-codes the `Accept-Language` header to `en-US,en;q=0.5`, which imposes a specific language/locale preference on all requests. This is a natural-language policy concern because the user is not offered a locale choice or informed that English is being forced.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.