ClawHub
Back to skill

Security audit

Alibaba Cloud AI Audio CosyVoice Voice Clone

Security checks for vulnerabilities and agentic risk

Overview

The skill is coherent and not deceptive, but it enables voice cloning with biometric identity implications without enough consent, retention, or deletion guidance.

Install only if you have rights and explicit consent to use the reference speaker's voice. Treat reference audio URLs, API responses, and voice_id values as sensitive, limit who can access them, and confirm how to delete custom voices in Alibaba Cloud when they are no longer needed.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding
The skill documents file read/write behavior via validation and output artifact steps, but it does not declare any tool scope such as permissions or allowed-tools. That mismatch can lead to overbroad or unexpected filesystem access at runtime, reducing auditability and making it harder for users or orchestrators to constrain what the skill may touch.

Session Persistence

Medium
Category
Rogue Agent
Content
# Model Studio CosyVoice Voice Clone

Use the CosyVoice voice enrollment API to create cloned voices from public reference audio.

## Critical model names
Confidence
82% confidence
Finding
The skill centers on creating cloned voices and reusing the returned voice_id in later TTS calls, which introduces persistent state tied to a biometric-derived identity artifact. If that identifier is retained, shared, or reused without controls, it can enable unauthorized voice impersonation, tracking across sessions, or misuse beyond the original user intent.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill instructs users to provide a public audio URL and names external Alibaba Cloud endpoints, but it does not clearly warn that reference audio and related metadata will be transmitted to third-party services. Because voice samples are biometric and potentially sensitive personal data, omission of an explicit disclosure can cause privacy, compliance, and consent failures.

Natural-Language Policy Violations

Medium
Confidence
92% confidence
Finding
The script sets `--language-hint` to `zh` by default, which imposes a specific language/locale when the user does not explicitly choose one. This is a natural-language policy concern because the tool defaults to a fixed language rather than offering a neutral default or requiring opt-in.

Static analysis

No suspicious patterns detected.