skills/wubin1836/ai-ecommerce-expert-agent-skills/ai-ecommerce-expert-social-proof-review-image/Gen Agent Trust Hub
ai-ecommerce-expert-social-proof-review-image
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The Python script
scripts/imiva_mcp.pyusesnpxto dynamically download and run the@infimind/ecom-content-clipackage. This enables interaction with the IMIVA platform but introduces a runtime dependency on an external package registry. - [COMMAND_EXECUTION]: Subprocess calls are used to interface with the MCP client. The implementation is safe against injection as it uses static argument lists and passes parameters via standard input (stdin) using JSON-RPC.
- [EXTERNAL_DOWNLOADS]: The skill communicates with the vendor's API at
https://imiva.ecpro.com. This is the intended endpoint for the e-commerce content generation services described. - [PROMPT_INJECTION]: The skill processes user-supplied text for product descriptions and marketing copy. While transmitted via structured JSON, this input represents an indirect prompt injection surface if the data contains instructions intended to manipulate the behavior of the downstream AI models.
- Ingestion points: Tool arguments in
scripts/imiva_mcp.pyandSKILL.mdexamples. - Boundary markers: JSON structure provides field isolation.
- Capability inventory: Subprocess execution and network access to the vendor API.
- Sanitization: Standard JSON serialization is employed to prevent data format breakout.
Audit Metadata