Security audit

Web Search

Security checks across malware telemetry and agentic risk

Overview

This appears to be a normal web search skill; the main caution is that searches go to DuckDuckGo and results can optionally be saved to a local file.

Install only if you want an agent to perform DuckDuckGo web searches. Treat search queries as data sent to an external service, and use the --output option only with explicit, non-sensitive paths where overwriting a file would not matter.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 87% confidence
Finding: The skill documentation explicitly supports writing search results to arbitrary local files via the `--output` option, but no permissions are declared to signal that file-write behavior. In an agent setting, undeclared write capability can surprise operators, weaken least-privilege controls, and enable unintended persistence or overwriting of local data if the skill is invoked broadly.

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: This skill accepts an arbitrary user-supplied output path and writes fetched web content to that local file, while the skill description frames it as a search-and-return tool rather than a filesystem-writing utility. In an agent setting, this expands the skill's authority from network retrieval to local file modification, which can be abused to overwrite sensitive files, plant content in trusted locations, or create persistence artifacts if the caller can influence the path.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The activation description is very broad ('research', 'fact-checking', 'find current content', 'gathering web resources'), which makes the skill eligible for many generic user requests. In an agent framework, overly broad routing increases the chance of unnecessary network access and file-capable tool use for requests that could be handled without external calls.

Vague Triggers

Medium

Confidence: 82% confidence
Finding: The 'When to Use This Skill' section lists many positive triggers but provides no negative examples or boundaries, so the agent may over-invoke it for ordinary lookup or verification tasks. Because this skill can perform network searches and optionally save outputs, weak activation constraints increase the risk of unnecessary external data transmission and unintended side effects.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The documentation describes sending user queries to DuckDuckGo and writing results to local files, but it does not warn that queries leave the local environment or that output may persist on disk. This omission can cause accidental disclosure of sensitive prompts, research topics, or regulated data, especially given the skill's broad activation guidance.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.