thought-mining

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

• Third-party content exposure detected (high risk: 0.80). The skill's required workflow (stages/03-validation.md) explicitly instructs the agent to perform "联网搜索" (internet searches), ingest and record web reference links from public sources to validate viewpoints, which means it will fetch and interpret untrusted third-party content that can materially influence decisions.