ClawHub
Back to skill

Security audit

Skill Auditor in Sandbox

Security checks across malware telemetry and agentic risk

Overview

This skill mostly does what it claims, but its audit script can collect and print more skill file contents than the selected target, so it needs careful review before use.

Use this only in a fresh disposable NovitaClaw sandbox with no unrelated installed skills or secrets. Expect reports to contain raw source text, and prefer waiting for a version that limits scans to the requested skill directory and redacts likely secrets.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

High
Confidence
98% confidence
Finding
The script enumerates files from the entire skills directory and exports their full contents, rather than limiting analysis to the requested SKILL_NAME. In a sandbox that contains multiple installed skills, this enables unnecessary bulk collection and disclosure of unrelated code and potentially embedded secrets, exceeding the stated purpose of auditing a single skill.

Context-Inappropriate Capability

Medium
Confidence
96% confidence
Finding
Collecting and outputting complete source contents is broader than required for a security audit report and creates a data-exfiltration primitive. Even in a sandbox, full source export can expose proprietary code, embedded credentials, or sensitive prompts contained in installed skills.

Missing User Warnings

High
Confidence
97% confidence
Finding
The script bulk-reads text files and exports their contents without clear user disclosure, creating an overbroad collection channel. Because the code currently scans the whole skills directory, this can leak unrelated installed skills, secrets, or proprietary content beyond the intended audit target.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.