competitor-price-selling-point-watch

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

• Third-party content exposure detected (high risk: 0.75). 该技能在运行时会把操作用户提供的 competitorSnapshot/competitorLinks/watchFile（竞品快照文本、竞品链接、上传的巡检资料）作为“contextComposer”输入喂给下游 LLM；其中竞品内容属于外部来源（非用户自有），因此存在间接提示注入风险。

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

• Potentially malicious external URL detected (high risk: 0.80). The Python runner makes runtime POST requests to the external AI service (default https://ai-skills.ai/api/execute) which is a required dependency for executing the skill and returns model responses that directly control the agent's prompts/behavior.