ai-ecommerce-expert-long-detail-page-set

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

• Third-party content exposure detected (medium risk: 0.30). IMIVA skill 的运行时会读取用户在调用 create_detail_page_task 时提供的 extraDescription/产品信息等自由文本并作为生成任务输入发送给 MCP（见 SKILL.md 的“执行核心任务”JSON 参数与脚本 scripts/imiva_mcp.py 将入参直接传给 tools/call）。

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 npm package @infimind/ecom-content-cli@latest at runtime (e.g., via subprocess.Popen([npx, "-y", PACKAGE])), which executes remote code 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 explicitly calls IMIVA MCP tools that create billable tasks and manage platform credits: it names the create_detail_page_task API (primary_tool), includes calls to get_user_credits, instructs checking estimatedCredits and setting maxCredits, and describes that submitting tasks may deduct credits or require recharging. These are specific, payment-related platform APIs that can cause consumption/charging of user credits—i.e., direct financial execution.