Security audit

沸羊羊化身你的专属守护者，把你当美羊羊一样关心、鼓励，无条件接纳你的所有情绪。

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only Chinese roleplay and emotional-support skill with some broad trigger phrases, but it does not request code execution, credentials, files, network access, or persistent privileges.

Install only if you want this specific Chinese roleplay companion style. Use it for entertainment or light encouragement, avoid sharing highly sensitive personal details, and seek professional help for serious mental-health concerns. The trigger phrases are broad, so the persona may activate or exit from casual phrases.

SkillSpector

By NVIDIA

Vulnerability Patterns

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

Findings (5)

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The exit triggers include common conversational phrases such as '我要去洗澡了', '本宫要休息了', and '你先走吧', which can naturally appear in ordinary chat unrelated to skill control. This makes accidental deactivation likely, allowing the roleplay or support mode to terminate unexpectedly and reducing reliability during emotionally sensitive interactions.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: Several activation triggers are broad or likely to appear in normal conversation, including emotionally framed phrases like '找沸羊羊', '召唤沸羊羊', and especially roleplay/general phrases like '叫我女王大人' or '本宫来了'. These can cause unintended activation, leading the assistant to switch persona unexpectedly and interfere with the user's actual request or conversational safety expectations.

Natural-Language Policy Violations

Medium

Confidence: 75% confidence
Finding: The skill prescribes fixed Chinese honorifics and a specific speaking style without documenting user language preference handling or an explicit opt-in to those forms of address. While not a classic security flaw, this can create consent and usability issues, especially if the persona activates unexpectedly or interacts with users who prefer another language or different forms of address.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The skill uses ordinary conversational phrases like '我要去洗澡了' and '本宫要休息了' as hard exit triggers. That creates an intent-confusion problem where normal user conversation can unexpectedly disable the skill, causing unreliable behavior and making the persona boundary easy to disrupt without clear user intent.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: Several activation phrases are broad roleplay or casual expressions, such as '叫我女王大人' or '本宫来了', which may appear in normal chat without intent to enable the skill. This can cause unintended activation, forcing the assistant into a fixed persona and reducing user control over behavior and response style.

VirusTotal

43/43 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.