Security audit

Free Web Search Js

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed web-search and page-fetching skill; it sends searches and URLs to external services and can render pages in Chromium, but I found no hidden exfiltration, credential access, or destructive behavior.

Install only if you are comfortable with your search terms, network region/IP metadata, and fetched URLs being visible to search engines, IP-check services, and destination sites. Avoid secrets, private internal URLs, or confidential research terms; use region overrides and no-fetch/http-only options when you want less automatic browsing.

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)

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The README advertises automatic fetching of top search results and describes IP/region detection, but it does not clearly warn users that queries, target URLs, and IP metadata will be sent to multiple third-party services. In a search/fetching skill, this can expose sensitive prompts, internal research topics, or operator network information through outbound requests users may not realize are automatic.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The trigger keywords are extremely broad ('search', 'find', 'look up', and common Chinese equivalents), making accidental activation likely during ordinary conversations. In a skill that performs live web queries and content fetching, unintended invocation can leak user prompts, sensitive terms, or internal URLs to third-party search engines and websites without sufficiently explicit intent.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The top-level description markets the skill as convenient and keyless but does not immediately disclose that user queries, fetched URLs, and retrieved page content are sent to and collected from third-party services. Because the core function is web search and page retrieval, this omission undermines informed consent and increases the chance that users submit sensitive or private data believing the action is local or low-risk.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The Playwright fallback loads arbitrary user-provided URLs in a full browser context, which executes remote page JavaScript automatically. In an agent/automation context, this increases attack surface for tracking, browser fingerprinting, and exploitation of browser/runtime vulnerabilities, especially because there is no allowlist, scheme restriction beyond a loose 'http' prefix check, or prominent warning before active content execution.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The skill automatically sends user queries to third-party IP-geolocation services and search engines, and later auto-fetches top result URLs, without an explicit consent or warning boundary. In this skill context, that can leak sensitive prompts, internal research topics, or agent-originating traffic metadata to external services and also causes the agent to visit attacker-controlled pages returned in search results.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.