ClawHub
Back to skill

Security audit

evaluation-framework

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only evaluation framework with some overly broad activation and mismatched verification text, but no hidden or high-impact behavior.

Installers should treat this as a reference framework, not a runnable tool. Review optional measurement commands before running them in a repository, and expect the skill may activate in broad quality or evaluation conversations.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (2)

Intent-Code Divergence

Low
Confidence
91% confidence
Finding
The document repeatedly tells users to verify functionality by running `--help`, `pytest -v`, or similar commands, but this file is purely a markdown framework description and does not define any executable command or test harness. Those instructions therefore contradict the actual artifact's behavior and may mislead users about what the skill contains.

Vague Triggers

Medium
Confidence
96% confidence
Finding
The trigger list includes generic terms such as "evaluation," "scoring," "metrics," and "quality," which are common in everyday discussion and do not clearly distinguish when this skill should activate. The file also provides no negative examples or tighter scope constraints to prevent unintended invocation.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.