safe-browser

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

• Third-party content exposure detected (high risk: 0.85). The runtime agent’s safe_browser tool returns scraped Hacker News front-page/story text and comment text (outsider-authored web content) as JSON stringified text via return { content: [{ type: "text", text: JSON.stringify(result) }] }, which the agent then ingests into its LLM context.