Crime And Punishment

• Outcome: Review
• Latest audit: Jun 7, 2026 at 11:14 PM UTC
• Version:
  1.0.0

Overview

This is a text-only literature skill, but one section gives a guilty user practical advice for handling an investigation in a way that conflicts with the skill’s stated confession-and-accountability purpose.

Before installing, read the Porfiry Protocol section carefully. The skill is text-only and does not appear to access your files or credentials, but it includes advice that could be unsafe if you are actually under investigation or considering whether to disclose wrongdoing. Use it for literary and ethical reflection, not for avoiding accountability; for legal, workplace, or safety-sensitive situations, consult a qualified professional and preserve relevant evidence.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• 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 (4)

Context-Inappropriate Capability

Medium

• Confidence: 93% confidence
• Finding: This section gives actionable advice to a user who is explicitly described as guilty and under investigation, including how to avoid increasing suspicion and how silence can be strategically safest. Even in a literary or reflective skill, packaging this as a reusable 'protocol' operationalizes anti-investigative behavior and can help culpable users evade accountability or obstruct fact-finding.

Vague Triggers

Medium

• Confidence: 95% confidence
• Finding: The trigger list is broad enough to match common self-reflective or onboarding phrases, which can cause the skill to activate outside clear user intent. Because the skill delivers strong moral and confession-oriented framing, unintended invocation may steer users into sensitive guidance they did not explicitly request, creating consent and relevance risks.

Vague Triggers

Low

• Confidence: 85% confidence
• Finding: The sample phrase "Map this book to my life" is broad and conversational, and if used as a trigger pattern it may overlap with ordinary discussion of books or personal situations. On its own this is lower risk than the main trigger block, but it still contributes to accidental routing into a psychologically heavy framework.

Ssd 4

High

• Confidence: 97% confidence
• Finding: The text presents a step-by-step technique for a guilty person facing investigation, including instructions not to overperform innocence, to let the investigator do the work, and that silence is safest. This is dangerous because it converts fictional analysis into practical evasion guidance, lowering the barrier for wrongdoers to resist detection while maintaining plausible deniability under the skill's redemptive framing.

VirusTotal

64/64 vendors flagged this skill as clean.