huashu-nuwa

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). Most links are to benign news, docs and academic sites, but multiple GitHub repositories from a single, unvetted personal account (alchaincyf) are being recommended for one‑line install/clone (i.e. pulling and running third‑party code), which is a security risk because installing unreviewed code can execute malware or supply‑chain attacks.

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

• Third-party content exposure detected (high risk: 0.85). Required runtime workflow includes “纯网络搜索/网络搜索补充”，通过 WebSearch/浏览器/抓取工具把公共网页、社交媒体、播客/访谈转录等“外部免费文本”读入并写入 research markdown，然后用于后续综合进入 LLM 上下文，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: 0.90). This repo explicitly instructs agents/runtimes to fetch and install the skill from https://github.com/alchaincyf/nuwa-skill (e.g., "帮我安装这个 skill: https://github.com/alchaincyf/nuwa-skill" and "git clone https://github.com/alchaincyf/nuwa-skill"), which at runtime pulls remote code/content that directly defines the skill's prompts/behavior and therefore can execute or control agent instructions.