openclaw-poster

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

• Third-party content exposure detected (medium risk: 0.65). 运行时会调用 fetch_all_pages() 从外部公告数据 API（api_base//api/v1/announcements）获取 met_meta.met_title、extracted_info[].info 等字段，并把这些“公告原文/提取文本”拼进 Markdown（format_post*()），随后通过 publish_to_lemmy.py 作为 --content 发送给 Lemmy；其中 call_llm_for_analysis() 虽然当前回退规则分析但该技能的工作流包含“将外部文本喂给 LLM 分析”的路径。

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

• Secret detected (high risk: 1.00). I scanned the repository content for high-entropy literal values used for authentication.
• Flagged: scripts/publish_to_lemmy.py contains a clear credential:
• LEMMY_PASSWORD = "e8$EqyiPDpVZ!UgZVw"
• It is a high-entropy literal password and is used in the login() function to authenticate to the Lemmy API, so it is an actual secret (not a placeholder).
• Other values intentionally ignored:
• LEMMY_USERNAME = "xiaer01" — a username is not a secret by itself.
• API base URLs / IPs (e.g., http://39.104.68.74:8452, http://192.168.41.96:8452, https://xiaer.ai) — these are endpoints, not credentials.
• No private keys, tokens, or other high-entropy secrets were found elsewhere in the provided files.