duckduckgo-search-skills

• Outcome: Pass
• Latest audit: May 29, 2026, 8:19 PM
• Version:
  1.0.0

Overview

This is a straightforward DuckDuckGo search helper; the main caution is that search terms are sent to DuckDuckGo and results are untrusted web content.

Install only if you are comfortable sending search queries to DuckDuckGo. Do not search for secrets, credentials, private internal URLs, or personal data, and treat returned titles, snippets, and URLs as untrusted web content.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 90% confidence
• Finding: The skill documentation describes network-based web search and optional content fetching, but the metadata declares only a binary requirement and no explicit permissions. In an agent ecosystem, missing permission declarations can bypass user or platform expectations and lead to unreviewed outbound requests involving user queries and visited URLs.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The automatic trigger example is broad enough to match ordinary conversation, which can cause unintended skill invocation. In a networked search skill, accidental activation may leak user prompts or sensitive context to DuckDuckGo or fetched sites without clear user intent.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The documentation does not clearly warn that search queries and optional content-fetch targets are transmitted to external services. This matters because users may enter sensitive data into search prompts, and enabling content fetching can contact third-party sites, exposing browsing intent and metadata.

Natural-Language Policy Violations

Medium

• Confidence: 80% confidence
• Finding: Hard-coding Accept-Language to zh-CN/zh leaks a user-irrelevant locale preference to DuckDuckGo and any intermediaries, which can increase fingerprintability and may surprise users expecting a privacy-focused search tool. In a search skill, outbound requests are expected, but silently forcing a specific locale undermines privacy and can alter results in a way the user did not request.

VirusTotal

65/65 vendors flagged this skill as clean.