quick-start

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md 的运行流程会在运行时通过 scripts/cache_docs.py 按 permalink 读取/缓存文档内容（references/.cache/... 或 ~/.cache/...），而这些文档来源于外部 WebFetch URL（如 https://jeesite.com/docs/...），属于“公共 web content fetched at runtime”并会以可读 Markdown 文本进入 LLM 上下文。

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill runtime invokes a cache script (python3 scripts/cache_docs.py) to fetch and load external documentation URLs (e.g. https://jeesite.com/docs/overview/) which would be injected into the agent's context and thus directly control prompts.