Back to skill

Security audit

chinese-voice-skill

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed Chinese text-to-speech helper, with the main caution being its unpinned user-level dependency install from a third-party package mirror.

Install only if you want Chinese TTS replies and are comfortable sending text to Microsoft TTS services. Prefer installing `edge-tts` in an isolated environment from a trusted, pinned source instead of running the documented mirror-based `pip install --user` command as-is.

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 (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:80
Finding
Unpinned Dependency Installation Through a Third-Party Package Mirror## Vulnerability Details **File Location**: `SKILL.md`, line 80 **Vulnerability Type**: Supply-chain risk caused by an unpinned dependency and an external package mirror **Risk Level**: Medium **Complete Code Snippet**: ```bash pip install --user edge-tts -i https://mirrors.aliyun.com/pypi/simple/ ``` ### Technical Analysis The Skill directs users or an executing agent to install `edge-tts` without specifying a reviewed version or verifying package hashes. It also overrides the default Python Package Index with a third-party mirror. Because the requested package version is mutable, a future installation may retrieve code that differs from the version available when the Skill was audited. A compromise of the upstream package, one of its transitive dependencies, or the configured mirror could cause malicious package content to be installed. Depending on the package distribution format and build process, package-controlled build logic may execute during installation. Installed malicious code could also execute when the TTS package is subsequently invoked. This is not evidence that the named package or mirror is currently malicious. The vulnerability is the absence of version and integrity controls around executable third-party dependencies. ### Attack Path 1. A user requests Chinese voice synthesis. 2. The environment does not already contain `edge-tts`. 3. The documented installation command is executed. 4. pip resolves the latest available package and transitive dependencies through the specified third-party mirror. 5. An attacker who has compromised the package, a dependency, or mirror-delivered artifacts supplies malicious package content. 6. Malicious logic executes during package build or installation, or later when the installed TTS component is invoked. ### Impact Assessment Successful exploitation could execute arbitrary code with the privileges of the account running pip or invoking the installed package. ...[truncated 360 chars]
Remediation
## Remediation Suggestions 1. Pin `edge-tts` and every transitive dependency to versions that have been reviewed and tested. 2. Record cryptographic hashes in a locked requirements file and install with hash enforcement: ```bash python -m pip install --require-hashes -r requirements.txt ``` 3. Prefer the official PyPI index unless a mirror is operationally required. If a mirror is required, document its trust model and use a controlled repository that verifies artifacts against approved hashes. 4. Install dependencies inside an isolated virtual environment rather than directly into the user's package directory. 5. Prepackage or provision the approved dependency during deployment instead of allowing runtime installation in response to a voice request. 6. Disable source builds where practical and accept only reviewed binary artifacts: ```bash python -m pip install --only-binary=:all: --require-hashes -r requirements.txt ``` 7. Run TTS processing with least privilege and restrict its access to credentials, sensitive files, and unnecessary network destinations.
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 (2)

Natural-Language Policy Violations

Medium
Confidence
89% confidence
Finding
The skill defaults to Chinese voice synthesis and states it will auto-trigger when voice replies are needed, but it does not indicate an explicit user choice, consent step, or language selection. In practice, this can produce unwanted audio output, leak user text to a remote TTS service, or create accessibility/usability issues when the user did not intend Chinese-language audio generation.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The trigger list is very broad and includes generic phrases like '开始语音', '开启语音', and '用语音回复', which can cause the skill to activate in situations where the user did not clearly request audio generation. Because the skill is configured to auto-trigger and can invoke a shell tool, unintended activation may lead to unexpected command execution, network access to Microsoft TTS services, and unsolicited media output.

Static analysis

No suspicious patterns detected.