xiaohongshu-title-score

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill tells the agent to prompt for or help set the REDFOX_API_KEY by inserting the actual key into shell/PowerShell commands (export/SetEnvironmentVariable) and to verify it with echo, which requires the LLM to handle/output the secret value verbatim.

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

• Third-party content exposure detected (medium risk: 0.30). 技能在运行时会调用脚本 scripts/fetch_xhs_trends.py 以用户给定的 keyword 向外部接口 https://redfox.hk/story/api/cozeSkill/getXhsCozeSkillData 拉取小红书爆款笔记的标题/作者/互动等文本并写入 关键词_爆款数据.md，随后 LLM 读取该文件进行标题分析与评分，因此存在“外部（第三方）作者文本被 LLM 摄入”的间接提示注入暴露面。