prd-to-tasks

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches and ingests external Feishu/Lark document URLs in Phase 1.0 ("Lark 文档调研" via lark-cli docs +fetch --doc "<referenced-url>") and loads user-provided KB markdown files, and those fetched documents are parsed to validate PRD assumptions and drive risk classification and routing decisions—meaning untrusted, user-generated third‑party content is read and can materially change subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill explicitly fetches Feishu/Lark document URLs (e.g., feishu.cn/docx/ and feishu.cn/wiki/) at runtime via the command lark-cli docs +fetch --doc "<url>" and ingests that remote document content into the agent context to drive its assessments and subsequent prompts, so these external URLs can directly control agent instructions.