ecommerce-review-root-cause-analysis

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context via the required runtime input fields reviewText (user-pasted差评/反馈) or reviewFile (uploaded评论/客服/售后文档) which are not authored by the operating user and are ingested as readable text by the skill’s context composer.

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

• Potentially malicious external URL detected (high risk: 0.90). The provided runner (scripts/run.py) makes runtime requests to the remote execution endpoint https://ai-skills.ai/api/execute (via DEFAULT_BASE_URL https://ai-skills.ai), which executes the skill/LLM and returns outputs that directly control the agent's prompts and behavior.