Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed mxai image/video generation integration that uses an API key and sends prompts or images to mxai for the requested creative tasks.
Install only if you intend to use mxai's external service. Prompts, uploaded images, image URLs, task history queries, and account-credit checks may be sent to mxai using your MX_AI_API_KEY, so avoid submitting private, regulated, or sensitive content unless you accept that handling. Be careful before setting the MJ public option to true because it may expose generated works beyond your private session.
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
- **Windows PowerShell**: $env:MX_AI_API_KEY="nb_xxx"
- **Windows CMD**: set MX_AI_API_KEY=nb_xxx
- **Docker**: -e MX_AI_API_KEY=nb_xxx
- **.env 文件**: 写入 MX_AI_API_KEY=nb_xxx
⚠️ 任何能设置环境变量的方式都可以，不依赖特定平台。
The skill routes user prompts and potentially uploaded/base64 images to a third-party HTTP API but does not present a prominent privacy notice at the point of use. Users may unknowingly send sensitive images or descriptive content off-platform, creating confidentiality and compliance risk if personal, proprietary, or regulated data is included.
The skill includes broad natural-language trigger examples like checking recent works or prior images/videos, which can overlap with ordinary conversation and cause the agent to invoke external HTTP actions unexpectedly. In a skill that submits generation jobs and queries account/task history, unintended activation can expose user prompts, images, or account metadata to the third-party service without sufficiently explicit user intent.
The MJ draw interface exposes a 'public' parameter that can make generated works publicly visible, yet the documentation does not give a conspicuous warning about the disclosure consequences. This can lead users to unintentionally publish sensitive prompts, reference images, or generated content beyond their expected audience.
The script's user-facing comments and output are entirely in Chinese, including the instruction to say a Chinese phrase to reload the skill. This imposes a specific language/locale on users without any opt-in or alternative, which matches the language/locale policy violation criteria.
No suspicious patterns detected.