wechat-write

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

• Third-party content exposure detected (medium risk: 0.30). 脚本 scripts/fetch_gzh_trends.py 在 Step 1/2 的“查询爆款→分析规律”路径上通过 POST https://redfox.hk/story/api/gzh/search/hotArticleNew 读取红狐接口返回的公众号文章字段（如 title/summary/url 与作者/互动数据），这些字段属于外部（第三方）作者文本并会被后续分析/生成文案所消化。

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

• Secret detected (high risk: 1.00). I found one literal, high-entropy API key in the repository that appears to be a real credential. It is not a placeholder or redaction and matches the kind of token format described in the policy (prefix + long hex-like string).

Ignored items and why:

• "ak_xxxx..." examples in SKILL.md and README files are placeholders (low-entropy / clearly templated) and therefore not flagged.
• Mentions of the environment variable name REDFOX_API_KEY and instructions to set it are just environment variable names / configuration guidance, not secrets.
• The script reading os.environ["REDFOX_API_KEY"] is usage of an env var (no secret hard-coded), so it is ignored.

Flagged because:

• references/gzh_trend_data_format.md contains a specific API Key string (ak_c4fc9018ffb14ce4ae35dafd92f466c3) in plain text, which appears to be a real, usable credential.