read-topic-collector

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md:5-7 and 145-167 describe using a runtime WebSearch to fetch and extract content from outsider-authored sources (e.g., Twitter/X, Product Hunt, Hacker News, Reddit, arXiv/blogs), which would then be ingested as readable text into the agent’s LLM context for summarization.