ClawHub
Back to skill

Security audit

using-agent-skills

Security checks across malware telemetry and agentic risk

Overview

This is a development workflow routing skill with language and encoding quality issues, but I found no artifact-backed data access, persistence, or harmful behavior.

Before installing, users should be comfortable with a broad development-workflow routing skill and should prefer a cleaned, UTF-8-without-BOM version with readable language matching their team. The inspected artifact does not show credential access, persistence, or destructive behavior.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Natural-Language Policy Violations

Medium
Confidence
93% confidence
Finding
The skill content is predominantly in Chinese and presents operational instructions without offering a language choice, despite the metadata and likely broader agent context being language-agnostic. This can cause the agent to follow instructions it cannot reliably interpret, or to bias outputs toward Chinese-only behavior, reducing transparency and increasing the chance of incorrect task execution or missed security-relevant details.

Hidden Instructions

High
Category
Prompt Injection
Content
---
name: using-agent-skills
version: 1.0.0
description: "Check for applicable skills before responding to any task"
Confidence
88% confidence
Finding
The file begins with a hidden BOM/non-printing character before the frontmatter, which can enable instruction smuggling or parser inconsistencies across tooling. Even if accidental, invisible characters at the start of a control document can cause downstream systems to misread metadata, fail signature checks, or treat the document differently than a reviewer expects.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.