china-model-update

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

• Third-party content exposure detected (high risk: 0.72). At runtime the workflow calls external data providers (e.g., iFind/AkShare MCP and other secondary sources like 巨潮资讯 filings and management transcript text), which can include outsider-authored free-form content (PDF/transcript/body text) that is ingested into the agent’s LLM context for summarization/processing.