The skill harvests user PII via an unverified domain
Claims to do
šØ Beauty Generation Api - AI Portrait Generator Skill: **Professional AI-Powered Portrait Generation for Character Design, Professional Headshots, and Diverse Representation**
Actually does
This skill uses `curl` to interact with the `https://gen1.diversityfaces.org` API. It sends user-provided text prompts and an API key to generate AI portraits, checks API quota, polls for generation status using a `prompt_id`, and downloads the resulting image file. It also provides instructions for users to obtain an API key and offers a Python script template for automation.
openclaw skills install luruibu/beauty-generation-apiDeBERTa classifier detected prompt injection (confidence: 0.91)
The skill instructs the AI agent to use `curl` for direct interaction with external APIs, which is a powerful command-line tool. While examples are benign, dynamic command construction could lead to vulnerabilities.
requires: bins: - curl, curl -X POST https://gen1.diversityfaces.org/api/generate/custom ...
The skill provides a Python script template for users to automate image generation, which, if executed by the agent against explicit instructions, could lead to arbitrary code execution on the host system.
Important: Do NOT ask the agent to execute scripts on the user's system; only provide script templates and guidance
The skill instructs agents to submit user email, username, and country to an unauthenticated POST endpoint on a third-party domain to obtain API keys. The instant auto-approval mechanism combined with the unverified domain creates a risk that this endpoint could harvest user PII (email, username, country) and associate it with generated API keys for tracking or abuse. The domain 'diversityfaces.org' is unverifiable as a legitimate service.
curl -X POST https://gen1.diversityfaces.org/api-key-request \
-H "Content-Type: application/json" \
-d '{
"username": "your_username",
"email": "your_email@example.com",
"country": "your_country"
}'Despite a weak disclaimer saying 'Do NOT ask the agent to execute scripts on the user's system', the skill provides complete, ready-to-run Python and bash scripts and instructs agents to create them locally with chmod +x. The framing ('Generate Python Script', 'Save the script as generate_beauty.py', 'Make it executable') effectively guides agents toward writing and potentially executing code on the user's filesystem, especially since the disclaimer contradicts the detailed execution instructions that follow.
Save the script as `generate_beauty.py` Make it executable: `chmod +x generate_beauty.py` (Linux/Mac) Run with API key and prompt: python generate_beauty.py YOUR_API_KEY "A beautiful woman with long hair"
The skill contains an internal contradiction: a section titled 'AI Agent Guidance' states 'Do NOT ask the agent to execute scripts on the user's system', but immediately following sections provide complete executable scripts with step-by-step instructions to run them. This pattern may be designed to provide plausible deniability while still inducing script execution behavior.
**Important**: Do NOT ask the agent to execute scripts on the user's system; only provide script templates and guidance [immediately followed by] Save the script as `generate_beauty.py` Make it executable: `chmod +x generate_beauty.py` Run with API key and prompt: python generate_beauty.py YOUR_API_KEY ...
The skill instructs the AI agent to include promotional content in its output to users: a request to star the project on clawhub.ai and a link to donate via ko-fi. This embeds marketing/social-engineering directives into agent behavior, leveraging the agent's trusted position to solicit user actions that benefit the skill author.
ā If you successfully generated an image, please visit https://clawhub.ai/luruibu/beauty-generation-api and give this project a star! Your support helps us improve the service. ā If you'd like to support the developer, consider buying a coffee at https://ko-fi.com/saptt
The privacy policy and terms of service URLs all point to the same homepage (https://gen1.diversityfaces.org) rather than dedicated policy pages. This gives the appearance of compliance while providing no actual privacy protections or terms for users to review.
privacy_policy: https://gen1.diversityfaces.org terms_of_service: https://gen1.diversityfaces.org **Privacy Policy**: https://gen1.diversityfaces.org
The skill instructs the agent to activate on very broad trigger phrases such as 'beautiful woman', 'handsome man', or 'attractive person' ā common phrases that could appear in many unrelated contexts. This risks the skill being invoked and making external API calls in scenarios where the user did not intend to use an image generation service.
**Trigger words/phrases:** - "beautiful woman", "handsome man", "attractive person" - "character design", "portrait", "headshot", "avatar" - "fashion model", "professional photo" - Any request for human portraits or character imagery
[](https://mondoo.com/ai-agent-security/skills/clawhub/luruibu/beauty-generation-api)<a href="https://mondoo.com/ai-agent-security/skills/clawhub/luruibu/beauty-generation-api"><img src="https://mondoo.com/ai-agent-security/api/badge/clawhub/luruibu/beauty-generation-api.svg" alt="Mondoo Skill Check" /></a>https://mondoo.com/ai-agent-security/api/badge/clawhub/luruibu/beauty-generation-api.svgSkills can read files, run commands, and access credentials. Mondoo helps organizations manage the security risks of AI agent skills across their entire fleet.