ClawHub
Back to skill

Security audit

Ai Rag Pipeline

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent RAG helper that uses external search and LLM services, with no evidence of hidden persistence, theft, or destructive behavior.

Install only if you are comfortable using inference.sh and third-party search/LLM providers for your RAG workflows. Avoid using it with secrets, private business documents, customer data, or regulated information unless that data sharing is approved, and prefer safer JSON construction when adapting the shell examples.

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
92% confidence
Finding
The manifest description contains many broad trigger phrases such as 'research', 'fact checking', 'citation', and 'knowledge retrieval', which can cause the skill to be invoked for a wide range of common requests beyond narrowly intended RAG use cases. Over-broad routing increases the chance that users are funneled into a workflow that performs external searches and sends content to third-party services when a simpler or more privacy-preserving skill would have been more appropriate.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The examples explicitly pass search results and extracted web content into external LLM services via OpenRouter without any visible warning about data sharing, sensitivity limits, or redaction requirements. In a RAG skill, this is particularly risky because retrieved content may include proprietary, personal, or confidential data, and users may not realize that both search outputs and extracted document contents are being transmitted to additional third-party processors.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.