notte-browser

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

• Third-party content exposure detected (high risk: 1.00). This skill explicitly navigates to and scrapes arbitrary public websites (e.g., notte page goto / notte page scrape in SKILL.md and examples like scraping https://news.ycombinator.com) and also starts AI agents that ingest page content (notte agents start), and the SKILL.md Security Notes explicitly state that scraped page text and agent observations are untrusted input that can contain prompt-injection attempts.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The notte CLI fetches arbitrary web pages at runtime (e.g., the example URL https://news.ycombinator.com via notte page goto / notte page scrape), and the documentation expressly states scraped page content is ingested into agent/tool context — a clear path for remote content to control prompts (prompt‑injection) during execution.