The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and execute the @infimind/ecom-content-cli package from the npm registry. This is the official command-line interface for the platform vendor (Infimind) and represents a legitimate vendor-owned resource for the skill's primary function.
[COMMAND_EXECUTION]: The helper script scripts/imiva_mcp.py invokes subprocess.Popen to launch the platform's MCP client. The implementation is secure as it uses a list-based command structure with a hardcoded package name and verified paths, preventing shell injection vulnerabilities.
[REMOTE_CODE_EXECUTION]: The skill executes the official IMIVA tool via npx to interact with remote content generation services. This execution is scoped to the vendor's infrastructure and is required for the e-commerce image generation logic.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided product data which is a standard surface for indirect injection. The instructions mitigate this by directing the agent to strictly follow verified facts and seller-confirmed parameters.
Ingestion points: Product images and descriptive text provided to the create_detail_page_task tool.
Boundary markers: Instructions include explicit warnings to "仅使用已核实卖点和参数" (only use verified selling points and parameters) and to perform budget checks before task creation.
Capability inventory: The skill allows creating and querying commercial content generation tasks via the vendor's API.
Sanitization: Input is passed through structured JSON-RPC payloads, which reduces the risk of control character manipulation during the bridge between the AI and the local system tools.

ai-ecommerce-expert-product-comparison-image