ClawHub
Back to skill

Security audit

Skill Quality Scorer

Security checks across malware telemetry and agentic risk

Overview

This skill is a benign skill-quality auditor that reads target skill files and runs a local read-only audit script when the user asks for skill scoring or review.

Install this if you want automated scoring of Agent Skill packages. Before using batch mode, point it only at skill directories you intend to inspect, because it will read SKILL.md files and related references/scripts under those targets.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding
The skill instructs the agent to read files and invoke a shell command (`python scripts/static_audit.py`) but does not declare any permissions or safety boundaries. This creates a transparency and control gap: callers may not realize the skill can access local files or execute commands, which increases the risk of unintended file exposure or command execution in sensitive environments.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The skill includes broad natural-language trigger phrases like generic requests to score or review a skill, which can cause accidental activation outside the user's precise intent. Because this skill can read files and invoke a shell-backed audit script, unintended triggering increases the chance of unnecessary file access or command execution on arbitrary directories.

Vague Triggers

Medium
Confidence
88% confidence
Finding
The sample trigger list includes a broad phrase (e.g. a generic request to check whether a SKILL.md 'meets standards') that can cause the skill to activate for unrelated review or compliance tasks. Over-broad triggers are dangerous because they increase accidental invocation, prompt-space collisions, and the chance that this skill overrides a more appropriate, narrower skill.

Natural-Language Policy Violations

Medium
Confidence
82% confidence
Finding
The embedded evidence strings are predominantly Chinese and suggest a fixed-language output style without any indication that language should follow user preference or locale. This is not a direct security exploit, but it can degrade safe operation by causing incorrect or inaccessible outputs, especially in evaluation workflows where language mismatch may hide warnings or produce misleading audit results.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.