ai-ecommerce-expert-reference-image-video

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). 在脚本 scripts/imiva_mcp.py 的运行时调用链中，AI 输入（如 --args 里的 prompt/referenceImages 等用户提交的 JSON 字段）会被直接打包为 MCP tools/call 参数并发送给 @infimind/ecom-content-cli@latest，因此工作流会在未先选取特定受信条目的情况下读取外部文本。

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes npx to fetch and run the remote package "@infimind/ecom-content-cli@latest" at runtime (subprocess Popen with [npx, "-y", PACKAGE]), which executes remote code that the skill depends on.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill exposes explicit MCP APIs that query and spend platform credits and can create billable tasks. It lists a primary tool create_video_generation_task, shows calling get_user_credits, using dryRun to obtain estimatedCredits, and creating tasks with idempotencyKey and maxCredits (which will consume credits). Those are specific, non-generic financial actions within the IMIVA billing/credits system and therefore constitute direct financial execution capability.