The Price of Inequality

• Outcome: Pass
• Latest audit: Jun 5, 2026 at 5:24 PM UTC
• Version:
  1.0.1

Overview

This is a non-executable educational skill about Stiglitz's inequality framework, with disclosed viewpoint bias but no hidden code, persistence, credential access, or data handling.

Install this if you want a Stiglitz-oriented explanation of inequality and policy responses. Be aware that it may activate on broad inequality-related topics and tends to present a progressive policy viewpoint, including advocacy language, so use other sources for balanced economic or political analysis.

SkillSpector

By NVIDIA

Vulnerability Patterns

• 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
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The trigger list activates on very generic terms such as “inequality,” “middle class,” and related broad economic concepts, which can cause the skill to hijack many unrelated conversations. Over-broad invocation is dangerous because it can override user intent, inject unsolicited ideological framing, and increase the chance that the assistant follows this skill when a different domain-specific response was intended.

Vague Triggers

Low

• Confidence: 82% confidence
• Finding: The onboarding prompt encourages users to say a broad educational phrase, “Teach me about inequality,” which overlaps with many general-purpose requests not necessarily intended for this specific skill. While less severe than the trigger metadata itself, it still promotes accidental routing into a narrowly framed skill and may bias responses when a neutral or broader treatment would be more appropriate.

Natural-Language Policy Violations

Medium

• Confidence: 97% confidence
• Finding: The content explicitly instructs users to support specific policies and to vote for inequality-reduction policies, which goes beyond informational summarization into direct political persuasion. In a voice/app skill context, this can improperly steer user political beliefs or behavior without balanced framing, user consent, or presentation of alternatives.

VirusTotal

64/64 vendors flagged this skill as clean.