wenext-3d-ordering

Fail

Audited by Gen Agent Trust Hub on Oct 19, 2026

Risk Level: HIGHDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill instructions reference and use sensitive configuration files located in ~/.config/wenext-3d-order/private.json. This file is explicitly described as containing recipient names, phone numbers, addresses, browser cookies, and OTP/SMS codes. Accessing these sensitive file paths and the credentials contained within them is a high-severity concern.
  • [COMMAND_EXECUTION]: The skill uses shell commands to invoke local Python and Bash scripts (order_assistant.py, quick_order_global.sh, quick_order_china.sh) to perform web automation via Chrome DevTools Protocol (CDP), file uploads, and order placement.
  • [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection.
  • Ingestion points: The skill ingests untrusted 3D model files (STL, STP, STEP) and processes dynamic content from the Wenext website via CDP automation.
  • Boundary markers: While 'Safety Rules' are present to guide the agent to stop at payment boundaries, these are natural language instructions and do not represent technical enforcement or data delimitation.
  • Capability inventory: The skill possesses significant capabilities including local script execution, writing logs and DOM snapshots to the file system, and automating web browser interactions.
  • Sanitization: No explicit sanitization or validation of the processed file data or retrieved web content is documented.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 19, 2026, 06:24 PM