soia-pkm-clip-x-profile

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

• Third-party content exposure detected (medium risk: 0.30). 该技能在网络模式下会读取由外部用户指定的 X 公开账号帖子正文与图片 ALT（通过 scripts/profile_x.py 调用 GET https://api.fxtwitter.com/2/profile/<handle>/statuses 获取 results，并在 normalize_status() 与 filter_records()/compile_prompt() 中把 status.text、alt_texts 等自由文本喂给后续摘要/Prompt Deck 编译），因此存在被外部作者注入恶意指令的间接提示注入风险。

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches profile data at runtime from https://api.fxtwitter.com (e.g., GET https://api.fxtwitter.com/2/profile//statuses) and directly uses the fetched post text/ALT to compile image prompts, so the remote content can control generated prompts.