ClawHub

LLM Eval Harness

Security checks across malware telemetry and agentic risk

Overview

This is a coherent LLM evaluation skill, but it tells agents to run generated code without defining safety boundaries.

Install only if you will run coding evaluations in a disposable sandbox or container with no sensitive credentials, minimal filesystem access, restricted network access, and explicit approval before executing any model-generated code.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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
Findings (1)

Unvalidated Output Injection

High
Category
Output Handling
Content
- Multiple judge passes for controversial cases

**Code execution:**
- For coding tasks, execute generated code against test cases
- Check for syntax errors, runtime errors, correct output

**Regex patterns:**
Confidence
92% confidence
Finding
execute generated code

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal