Security audit

deepsop-声音克隆助手

Security checks for vulnerabilities and agentic risk

Overview

This voice-cloning skill does what it claims, but it handles sensitive voice data and includes unsafe or under-disclosed setup and consent guidance that should be reviewed before use.

Install only if you are comfortable sending voice samples, synthesis text, and generated audio links to the AI Artist/OSS service. Use only voices you own or have explicit permission to clone, avoid impersonation or misleading content, rotate the exposed example key if it could ever have been valid, and store your own API key in a private environment or .env file that is not committed or shared.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (9)

Intent-Code Divergence

Medium

Confidence: 98% confidence
Finding: The documentation includes a real-looking API key value in example commands, which normalizes secret disclosure and may expose an actual credential if the token is genuine. Publishing credentials in setup instructions can lead to account abuse, billing loss, and downstream compromise of associated services.

Description-Behavior Mismatch

Medium

Confidence: 83% confidence
Finding: The skill can download remote synthesized audio and save it to a local workspace directory without that persistence behavior being disclosed in the manifest. Undisclosed local writes create privacy and data-handling risk, especially because generated speech may contain sensitive or user-identifying content and will persist on disk beyond the immediate request.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The README explicitly promotes uploading local audio and automatic upload to OSS/remote API endpoints, but it does not warn users that voice samples are sensitive biometric/personal data being transmitted to third-party services. In a voice-cloning skill, that omission materially increases privacy and consent risk because users may upload their own or others' recordings without understanding retention, processing, or authorization implications.

Missing User Warnings

Low

Confidence: 82% confidence
Finding: The README tells users to place the API key directly into shell environment variable commands but does not warn them to keep the credential secret or avoid exposing it in shared terminals, shell history, screenshots, or logs. This is a documentation security weakness rather than an exploit by itself, but it can lead to accidental credential disclosure and unauthorized API use.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill sends uploaded audio and synthesis text to an external API, but the documentation does not prominently warn users that their content leaves the local environment. This weakens informed consent and increases privacy risk, especially for voice samples, which are biometric/identity-linked data.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The examples explicitly mention cloning leaders' or celebrities' voices without any consent, impersonation, or abuse warning. In the context of a voice-cloning skill, omission of consent requirements materially increases the chance of fraud, social engineering, harassment, and reputational harm.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The documentation instructs users to upload local audio files to a third-party OSS-backed service but does not warn that the files may contain biometric voice data and could become remotely accessible via returned URLs. In a voice-cloning skill, this omission is more dangerous than a generic file upload because users may upload sensitive recordings whose privacy, consent, retention, and reuse implications are significant.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill uploads local audio to a third-party service for cloning without any built-in consent, ownership, or privacy warning. Voice samples are biometric-like identifiers and may belong to someone other than the operator, so silent upload and cloning materially increase impersonation, privacy, and compliance risk.

Ssd 3

High

Confidence: 99% confidence
Finding: Exposing a concrete API key-like value in setup examples is a sensitive data disclosure issue. Even if illustrative, it teaches unsafe secret handling; if real, it can enable unauthorized API usage, fraud, or service abuse.

Static analysis

No suspicious patterns detected.