ClawHub
Back to skill

Security audit

fusion-search

Security checks across malware telemetry and agentic risk

Overview

This is a real search skill, but it uses stealth browser scraping with broad auto-triggers and weakened browser protections, so users should review it before installing.

Install only if you are comfortable with automated web scraping through a stealth Playwright browser. Avoid sensitive searches, prefer specifying an engine when privacy matters, keep full-content fetching off unless needed, and run it in a contained environment because the Chromium launch weakens several browser protections.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
96% confidence
Finding
This script deliberately spoofs multiple browser fingerprint surfaces, including webdriver, plugins, languages, platform, hardware characteristics, permissions, canvas, and WebGL, to evade bot detection. In the context of a search skill, this goes beyond normal compatibility handling and materially enables covert scraping against sites that rely on anti-automation controls, increasing the risk of policy bypass and abusive collection.

Vague Triggers

High
Confidence
93% confidence
Finding
The trigger keywords are extremely broad, including common phrases like '搜索', '最新', '是什么', 'search', 'find', and 'research', which are likely to appear in normal conversation. This can cause the skill to activate unexpectedly and exfiltrate user prompts or context to external search providers without clear user intent, especially because the skill performs live network searches and optional full-page fetching.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The skill markets broad search and full-content retrieval but does not clearly warn that queries, selected URLs, and potentially fetched page contents are sent to third parties. In this context, the omission is meaningful because the tool supports multi-engine routing, fallback, and full-content scraping, increasing the amount of user-derived data transmitted externally.

Natural-Language Policy Violations

Medium
Confidence
88% confidence
Finding
The script hard-codes navigator.languages to a specific Chinese-first profile without user consent, misrepresenting the user's locale and making requests appear to originate from a different audience segment. In this skill, that setting is part of a broader stealth/fingerprinting evasion pattern, so it can be used to manipulate content served by sites, bypass regional heuristics, or conceal the true client profile.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.