Web Search

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed web-search integration, but it relies on an external authenticated CLI and third-party search/extraction services.

Install only if you are comfortable trusting the inference.sh CLI and sending search terms, URLs, extracted page content, and example LLM prompts to external providers. Prefer the manual checksum verification path, review infsh commands before approving agent execution, and avoid submitting secrets, private internal URLs, regulated data, or proprietary content.

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 (2)

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The manifest description includes very broad trigger phrases such as "research," "search assistant," "agents," and "fact-checking," which can cause the skill to activate in ordinary conversations that do not clearly imply user consent to use external web services. In an agent setting, this increases the chance of unintended web requests and disclosure of user prompts or URLs to third-party providers.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The documentation explains how to send queries and URLs to Tavily/Exa via inference.sh but does not clearly warn that user-provided search terms, URLs, and possibly extracted content are transmitted to external services. This can lead to accidental disclosure of sensitive prompts, internal links, or proprietary data when users treat the skill like a local capability.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal