Murder Mystery Creator

• Outcome: Pass
• Latest audit: Jun 4, 2026, 4:08 PM
• Version:
  1.8.0

Overview

This is a coherent murder-mystery writing skill; the main cautions are mature crime themes, automatic asset creation, and a generated HTML page that loads D3 from a CDN.

Install only if you want a skill that can generate full murder-mystery packages, including mature crime themes and many local output files. Review generated HTML before sharing it, and consider replacing the CDN D3 dependency with a local or pinned copy if offline use or supply-chain control matters.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (7)

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The template loads D3 from a third-party CDN at runtime, creating a supply-chain and privacy risk. If the CDN content is tampered with, blocked, or replaced, the page executes untrusted JavaScript in the user's browser; this is especially relevant because the template later renders large amounts of generated story data into the DOM.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The skill advertises broad trigger phrases such as requests to 'make a murder mystery' or 'analyze a work then make one,' which can overlap with ordinary conversation and cause unintended activation. In an agent setting, this can route unrelated user input into a long, stateful generation workflow, producing unexpected behavior, wasted tool usage, or accidental disclosure of intermediate reasoning/templates.

Vague Triggers

Medium

• Confidence: 93% confidence
• Finding: The Mode B trigger rule ('mention a specific work + request generation/learning/reference') is underspecified and lacks scope limits, allowing casual discussion of media references to be misclassified as an instruction to analyze and imitate structure. This increases the chance of unintended skill execution and may also push the agent into copyright-sensitive or policy-sensitive transformation behavior without clear user intent.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The guide explicitly directs the system to automatically generate images and package outputs without any user confirmation or disclosure. In an agent setting, this can lead to unexpected file creation, resource consumption, and downstream output generation that the user did not knowingly authorize.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The workflow forces an automatic transition into a visual-material generation phase after script completion, followed by structured file packaging. This creates a non-consensual side effect in the agent workflow and increases the risk of surprise output creation, storage use, and unnecessary model/tool invocation.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The trigger phrase for Mode B is overly broad and can cause the skill to enter a high-action generation workflow when the user is only discussing a reference work, asking for critique, or making a casual mention. In this skill context, that misrouting can lead to unintended analysis or derivative content generation from named works without an explicit user confirmation boundary.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The file is a complete murder-mystery script centered on a gun homicide, deception, concealment of evidence, and player-directed manipulation tactics. In this context it appears intended as entertainment content rather than real-world criminal instruction, but the absence of any violence/mature-content warning or audience gating can expose users to unexpectedly graphic themes and normalize detailed violent scenarios, especially in a broadly accessible skill.

VirusTotal

62/62 vendors flagged this skill as clean.