Security audit

免费网页搜索工具

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly describes a web-search helper, but it requests raw command/read authority and contains broad generic automation claims that are not tightly scoped to search.

Install only if you are comfortable giving the agent command execution for package/browser setup and web search. Avoid using it for sensitive internal queries, keep full-text fetching off unless needed, and treat the generic file/API/command automation claims as out of scope unless you explicitly approve each action.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The documentation is internally contradictory about whether export operations are supported in the free version. In an agent setting, contradictory capability claims can cause the model to attempt unintended actions or misuse parameters such as input_params for unsupported operations, increasing the chance of unsafe tool invocation and user deception.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The skill is marketed as a narrow web-search tool, but later text broadens it into generic file handling, API integration, and command execution. This scope expansion is dangerous because agents may grant or exercise read/exec capabilities for tasks far beyond the user's expectation, creating a pathway to over-privileged behavior and unsafe command use.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The trigger condition is broad enough to activate for loosely related SEO and traffic-optimization requests without clearly limiting the skill to benign search use. Overbroad activation can cause the agent to invoke external browsing or exec-backed workflows unexpectedly, including in contexts where privacy, policy, or user intent do not justify it.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill does not clearly warn that search queries and optional full-text retrieval are transmitted to external search engines and target websites. This can expose sensitive prompts, research topics, or internal identifiers to third parties without informed user consent, especially when full-page fetching is enabled.

Natural-Language Policy Violations

Medium

Confidence: 90% confidence
Finding: The skill defaults to Chinese locale behavior and engine routing without a clear opt-in. Locale forcing can unintentionally steer user queries to region-specific services, alter result sets, and disclose geographic or language assumptions that the user did not choose, which is especially problematic for privacy-sensitive or compliance-sensitive searches.

Static analysis

No suspicious patterns detected.