trawl

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

• Third-party content exposure detected (medium risk: 0.30). Trawl’s runtime workflow is DISCOVER → CLASSIFY → DESIGN → COMPLY → DELIVER and, in its “extraction”/crawl pipeline specs, it assumes fetching and parsing HTML/near-duplicate detection on crawled page text and structured data (user-submitted content via external web pages), making the agent exposed to outsider-authored free text that can be injected through page content.