ai-ecommerce-expert-taobao-ecommerce-content

Pass

Audited by Gen Agent Trust Hub on Aug 25, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill executes the official @infimind/ecom-content-cli package from the npm registry using npx to perform content generation. This allows the agent to access IMIVA's external ecommerce tools.
  • [EXTERNAL_DOWNLOADS]: Fetches and runs the latest version of the official ecommerce content CLI from the IMIVA platform's registry at runtime.
  • [COMMAND_EXECUTION]: The scripts/imiva_mcp.py script utilizes subprocess.Popen to manage the lifecycle of the external CLI tool. It acts as a bridge for JSON-RPC communication between the agent and the product generator.
  • [CREDENTIALS_UNSAFE]: The skill manages authentication via an MCP_TOKEN environment variable. The implementation correctly scopes this credential to the execution environment of the service process.
  • [PROMPT_INJECTION]: The skill processes external assets such as productImages and user-defined generationStyle which are processed by AI models. While the skill uses structured JSON parameters, it acts as a surface for indirect prompt injection as it lacks explicit boundary markers to prevent models from potentially following instructions embedded within user-provided media or style descriptions.
  • Ingestion points: create_product_main_image_task arguments (file: SKILL.md, scripts/imiva_mcp.py).
  • Boundary markers: Absent.
  • Capability inventory: subprocess.Popen (file: scripts/imiva_mcp.py).
  • Sanitization: Uses JSON structure to organize parameters before passing to the CLI tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 25, 2026, 08:02 AM
Security Audit — agent-trust-hub — ai-ecommerce-expert-taobao-ecommerce-content