ClawHub
Back to skill

Security audit

个性化助眠脑波音乐定制系统

Security checks for vulnerabilities and agentic risk

Overview

This is a text-only Chinese sleep-audio recommendation guide with health-adjacent risks, but no hidden code, persistence, credentials, or unrelated system access.

Install only if you want a Chinese-language sleep-audio recommendation workflow. Use it with explicit user consent, do not treat it as medical diagnosis or a substitute for licensed care, and verify or replace the HTTP audio links with trusted HTTPS links before sharing them.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
97% confidence
Finding
The file distributes audio resources over plain HTTP, which allows network attackers to intercept or modify the linked content in transit. In a health-related skill that outputs personalized intervention links, this is more dangerous than a generic media catalog because users may trust the recommendation and consume tampered content presented as clinical guidance.

Vague Triggers

High
Confidence
96% confidence
Finding
The skill explicitly instructs immediate invocation for extremely broad, everyday sleep-related statements, including vague phrases like '最近睡眠不太好' and '想睡得好一点'. This can cause the agent to over-trigger into a quasi-medical screening and product-matching flow without sufficient user intent, appropriateness checks, or informed framing, increasing the risk of unwanted medicalized guidance and inappropriate recommendation of proprietary interventions.

Natural-Language Policy Violations

Medium
Confidence
72% confidence
Finding
The skill content is effectively hardcoded for Chinese-language interaction and presents a fixed Chinese diagnostic/questionnaire flow without offering language choice. In a general assistant environment, this can degrade user comprehension, impair informed consent, and increase the chance of misunderstanding during health-related questioning and recommendations.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The documentation provides symptom-based sleep-audio selection and severity mapping for insomnia-related conditions but contains no warning, limitation, or escalation guidance for medical red flags, vulnerable populations, or cases needing clinician review. In this context, the skill is positioned as a professional assessment and intervention tool, so omission of safety boundaries can lead to inappropriate health advice, delayed care, or overreliance on the audio plan as treatment.

Static analysis

No suspicious patterns detected.