Gang Leader For A Day

• Outcome: Review
• Latest audit: Jun 10, 2026 at 10:05 PM UTC
• Version:
  1.0.0

Overview

This is a documentation-only book skill, but it gives literal real-world fieldwork instructions for risky settings without adequate safety or ethics guardrails.

Install only if you want a book-analysis aid and are comfortable with sensitive content about gangs, poverty, policing, drugs, and race. Do not treat its fieldwork advice as operational guidance; real research in vulnerable or high-risk communities should involve institutional review, consent/privacy protections, local expertise, safety planning, and legal/ethical supervision.

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 (4)

Vague Triggers

High

• Confidence: 91% confidence
• Finding: The trigger rules are broad enough to activate on common topics like ethnography, economics, poverty, or users saying they just installed the skill. That can cause unsolicited insertion of gang/crime-related content into unrelated conversations, creating unsafe context switching and reducing user control over when the skill engages.

Vague Triggers

Medium

• Confidence: 83% confidence
• Finding: The phrase 'whenever I sense this book could help' is ambiguous and leaves invocation boundaries undefined. In practice, underspecified routing increases the chance of accidental activation and unrequested guidance on sensitive topics involving gangs, policing, and illegal economies.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The skill advises users to visit a field site repeatedly, avoid standard safeguards, and accept invitations in a potentially dangerous environment, without any safety, legal, IRB, or ethical warnings. Because the skill centers on gangs, housing projects, and illicit economies, a user could be harmed, trespass, compromise research ethics, or escalate risk by following this advice literally.

Natural-Language Policy Violations

Medium

• Confidence: 90% confidence
• Finding: The document reproduces a racial slur in quoted material without a clear content warning, editorial framing, or guidance on how the term should be handled. In an agent skill, this can normalize harmful language, lead the model to repeat the slur in user-facing outputs, and create reputational, safety, and harassment risks even though the surrounding context is literary/ethnographic rather than explicitly hateful.

VirusTotal

VirusTotal findings are pending for this skill version.