ClawHub

高情商聊天伴侣 AI Companion

Security checks across malware telemetry and agentic risk

Overview

This is a prompt-only social coaching and role-play skill with broad activation phrases, but no evidence of hidden access, persistence, exfiltration, or destructive behavior.

Install only if you want the assistant to handle a wide range of reply-writing, apology, comfort, role-play, and social communication requests. Review outputs carefully for workplace, relationship, or mental-health-adjacent situations, and be aware that generic requests like “How should I respond?” may activate this skill.

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 (4)

Vague Triggers

Medium
Confidence
91% confidence
Finding
The trigger phrases are broad, natural-language requests like '怎么回复比较好' and 'How to respond' that can appear in many ordinary conversations. This creates an activation-boundary problem where the skill may be invoked unintentionally, causing prompt hijacking of unrelated tasks or unexpected persona/roleplay behavior.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The English triggers are especially generic, including phrases such as 'Help me reply', 'Chat tips', and 'How to apologize', which overlap heavily with normal assistant usage. In a skill-routing environment, this can cause the skill to intercept common requests and steer responses into its prompt-defined behavior even when the user did not intend to use this skill.

Vague Triggers

Medium
Confidence
89% confidence
Finding
The intent-routing logic includes a broad catch-all for '其他情感相关 → 自动选择合适模式', which makes activation boundaries ambiguous and expands the skill beyond clearly defined intents. Such open-ended routing increases the chance of accidental invocation and makes it harder for the platform to predict when this skill will take control of a request.

Vague Triggers

Medium
Confidence
94% confidence
Finding
The quick-trigger list repeats many vague activators like '帮我回复', '怎么说比较好', 'What should I say', and 'How to respond' without scope constraints. This increases the probability that routine assistant prompts will match the skill and be rerouted into companion or script-generation behavior, which is a real routing security and reliability issue.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal