ai-legal-standard-v2

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md (Chapter 36.3 "持续合规监控 / 定期检查（每小时）
• 检查官方网站") explicitly requires the system to fetch and analyze public/official websites to identify new regulations and then automatically classify, analyze impact, update the knowledge base and push actions, so the agent ingests third‑party public web content that can materially influence subsequent tool use and decisions.

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

• Potentially malicious external URL detected (high risk: 1.00). The README includes a direct installation command to clone and run remote code from https://github.com/bytedance/deer-flow.git (git clone then pnpm install && pnpm dev), which fetches and executes external code and DeerFlow is described as a required core dependency for the skill.