wewrite-topic

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

• Third-party content exposure detected (high risk: 0.78). At runtime the skill can ingest outsider-authored free text via “热点抓取/ WebSearch” and “搜狗结果” in steps 2.1 and 2.1b, which then feed into LLM-based topic generation using the prompt rules in references/topic-selection.md (despite some being “degraded”/skipped on errors, when available they are parsed into JSON lists with title/description/urls/summaries/attributes).