Security audit

龙虾文游系统

Security checks across malware telemetry and agentic risk

Overview

This is a coherent text-adventure game skill with local save/load behavior and an optional LLM extension example, but users should avoid entering private information they would not want saved or sent to a model provider.

Reasonable to install for gameplay. Do not put passwords, private personal details, or confidential material into game scripts or actions if you may save the game. Only add the README's external LLM API example if you understand which provider receives the gameplay text and how that provider handles retained data.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Missing User Warnings

Low

Confidence: 91% confidence
Finding: The README advertises automatic save/load and identifies a local save directory, but does not warn users that gameplay content may be persisted on disk. In a text-adventure context, user-entered story content can include personal or sensitive information, so undisclosed local storage creates a privacy and data-handling transparency issue.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The LLM integration example sends current scene data, player state, and player actions to an external API without any privacy notice, minimization guidance, or consent language. Because freeform game input may contain personal data or sensitive roleplay content, this can result in unintended third-party disclosure.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The invocation examples are broad natural-language phrases such as ‘开始文字游戏，主题是[你的主题]’, which can overlap with ordinary conversation and cause accidental triggering. In a chat environment, this increases the risk that unrelated user messages are interpreted as skill activation, leading to unintended mode switching or handling of sensitive text as game content.

Vague Triggers

Low

Confidence: 93% confidence
Finding: The skill exposes very common command words—‘帮助’, ‘退出’, ‘状态’, ‘重新开始’, ‘存档’, ‘读档’—without any scope constraint tying them to an active game session. These generic tokens are likely to collide with ordinary user intent or platform-level commands, enabling accidental command execution and confusing or unsafe state changes.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.