smart-construction-analysis

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow performs outsider-ingested web browsing via web.search/web_fetch in Step 1–2 (“商业报道 + 展会 + 案例” followed by fetching), so attacker-authored free text from the open web can be read by the agent without preselecting a specific trusted item.