Back to skill

Security audit

VoxCPM2 — Tokenizer-Free Multilingual TTS

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent VoxCPM text-to-speech guide, but users should be careful with voice cloning, package installs, and local demo servers.

Install only in a virtual environment or container, prefer pinned versions or reviewed commits, and keep demo/API servers bound to localhost unless you add authentication and network restrictions. Use cloning or fine-tuning only with voices and transcripts you have rights and consent to use, and do not use it for impersonation or fraud.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (4)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:15
Finding

Unpinned VoxCPM Package Installation

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
SKILL.md:64
Finding

Execution of a Mutable VoxCPM Repository Revision

Content
View full analysis
Remediation
View remediation
git verify-commit pip install . ``` ]]>

T08 · Insecure Dependencies

Warning
Location
SKILL.md:72
Finding

Execution of a Mutable vLLM-Omni Repository Revision

Content
View full analysis
Remediation
View remediation
git verify-commit uv pip install . ``` ]]>

T08 · Insecure Dependencies

Warning
Location
SKILL.md:76
Finding

Unpinned Nano-vLLM VoxCPM Package Installation

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill provides explicit voice cloning and fine-tuning workflows using reference audio and transcripts without any warning about consent, impersonation, or privacy risks. In a TTS/voice-cloning skill, omission of these safeguards materially increases the chance of misuse for unauthorized cloning, social engineering, or processing personal voice data without permission.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 79)May include surrounding context.

uv pip install vllm==0.19.0 --torch-backend=auto git clone https://github.com/vllm-project/vllm-omni.git && cd vllm-omni && uv pip install -e . vllm serve openbmb/VoxCPM2 --omni --port 8000 curl http://localhost:8000/v1/audio/speech -H "Content-Type:application/json" -d '{"model":"openbmb/VoxCPM2","input":"Hello!","voice":"default"}' --output out.wav

text
Nano-vLLM: `pip install nano-vllm-voxcpm` (RTF ~0.13 vs standard ~0.30)

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The web demo and serving instructions start local HTTP services but do not warn that binding or reverse-proxying them beyond localhost can expose an unauthenticated speech API or demo UI. While the examples use localhost-oriented commands, users may deploy them on shared machines or servers and unintentionally expose model endpoints to other users or networks.

Content

No source excerpt is available for this finding.