travel-planner

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

• Third-party content exposure detected (medium risk: 0.30). 在 SKILL.md 的标准流程里需要从“用户发布的内容/帖子”中读取小红书信息并据此生成候选（目的地发现阶段），且该外部文本会被 LLM 作为行程信号使用。

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I reviewed all files for literal high-entropy secrets (API keys, private keys, long tokens). The only candidate I found is a long, random-looking query parameter value used as sealed_token in image URLs in README.md. It is high-entropy and appears to be an access token embedded in a URL (likely granting the ability to generate or view the star-history chart), so it meets the definition of a secret.

All other values are either documentation placeholders (e.g., AMAP_KEY, YOUR_API_KEY), short/simple strings (e.g., "openclaw", "password"-style examples), configuration names, URLs, or numeric/data values (coordinates, prices) which per the rules are ignored as non-secrets.