Nm Abstract Skills Eval

• Outcome: Review
• Latest audit: Jun 7, 2026 at 9:45 PM UTC
• Version:
  1.8.6

Overview

This skill is mostly coherent as a skill-auditing guide, but its testing guidance includes unsafe examples for executing tool paths from skills under review.

Install only if you are comfortable with a broad skill-evaluation assistant. When auditing unfamiliar skills, do not run generated integration or benchmarking code that executes declared tool paths unless you have reviewed the paths, constrained them to trusted locations, and run them in an isolated sandbox. Treat auto-fix, chmod, pip install, scan-all, and directory-wide commands as explicit user-directed maintenance actions, not automatic checks.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Output Handling: Unvalidated Output Injection, Cross-Context Output, Unbounded Output
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (2)

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The trigger list includes broad, common terms such as "evaluation," "improvement," and "skills," which are likely to appear in many unrelated conversations. This can cause unintended activation of the skill outside its intended auditing context, increasing the chance that the agent injects irrelevant instructions or tool-use guidance into unrelated tasks.

Unvalidated Output Injection

High

• Category: Output Handling

Content

# Test tool runs with --help
            try:
                subprocess.run([tool_path, '--help'],
                              capture_output=True,
                              timeout=5)
                results.tools_functional.append(tool)

• Confidence: 87% confidence
• Finding: subprocess.run([tool_path, '--help'], capture_output

VirusTotal

64/64 vendors flagged this skill as clean.