Algorithms Of Oppression

• Outcome: Pass
• Latest audit: Jun 10, 2026 at 1:48 AM UTC
• Version:
  1.0.0

Overview

This is a text-only educational skill about algorithmic bias, with no executable code, persistence, credential use, or hidden high-impact behavior.

Installers should treat this as an educational guide. The main practical risk is content exposure: some exercises ask users to search terms associated with examples of racist or sexualized search results, so proceed only if prepared to see disturbing material.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• 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 (2)

Missing User Warnings

Medium

• Confidence: 87% confidence
• Finding: The file instructs users to search terms such as "Black girls" and explains that the results included pornography, but it does not warn users beforehand that performing the exercise may surface explicit or harmful material. Because this is a markdown file and the behavior could affect user wellbeing or browsing safety, a brief user warning is expected.

Scope Creep

Low

• Category: Excessive Agency

Content

**How to do it:** Use DuckDuckGo, Ecosia, or other search engines that do not track users. Compare results. Different algorithms produce different biases.

**Case: The DuckDuckGo Test.** Noble tested her searches on alternative search engines. Some produced better results. Some were worse. The point: the problem is systemic, not limited to one company.

## Technique 3: Report and Document

• Confidence: 70% confidence
• Finding: not limited to

VirusTotal

64/64 vendors flagged this skill as clean.