wenqu-write

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). Designs that automatically copy plugin-provided "snapshot" files into local memory, persist potentially sensitive local secrets in agent memory, and unconditionally inject memory indexes into agent/system prompts create clear supply‑chain and data‑exfiltration risks (sensitive local data may be sent to external LLM/tools or introduced by third‑party snapshots).

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

• Third-party content exposure detected (medium risk: 0.30). 技能的运行流程在 Step 2（获取素材）会调用 wenqu-library 的联网搜索/下载/抓取并把抓取到的网页原文登记进 references/materials/，且该网页内容来自用户可影响的搜索结果或用户提供素材，属于外部/第三方作者自由文本被直接读入的路径。