linkfox-multimodal-extract-attributes
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill's primary function is to interact with the LinkFox multimodal AI API to extract visual attributes from product images. All operations are consistent with the documented purpose.
- [EXTERNAL_DOWNLOADS]: The
onboarding.pyscript identifies missing Python dependencies such asrequestsandqrcodeand instructs the user to install them via the standard package managerpip. This is a routine requirement for the script's functionality. - [DATA_EXFILTRATION]: The scripts transmit product data, phone numbers, and authentication tokens to vendor-controlled domains including
linkfox.com,tool-gateway.linkfox.com, andapi.linkfox.com. This is necessary for the skill's operation and follows the vendor's infrastructure layout. - [CREDENTIALS_UNSAFE]: The skill manages authentication via environment variables like
LINKFOX_AGENT_API_KEY. The onboarding process provides a path for users to securely obtain and configure these keys on their local systems. - [COMMAND_EXECUTION]: The skill provides instructions for the user to manually execute shell commands to persist environment variables in profile configuration files (e.g.,
.zshrc,.bashrc). These are standard administrative actions for setting up API-based tools. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from product data and user-provided prompts. While this constitutes an attack surface, the risk is considered low as the data is passed to a managed multimodal AI service for specific attribute extraction tasks.
Audit Metadata