katana-web-crawl

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

• Third-party content exposure detected (high risk: 0.75). Yes—Katana’s runtime crawling of outsider-provided web targets will ingest outsider-authored HTML/JS/response bodies and then the workflow instructs summarizing “notable route groups, JS endpoints, forms, XHR URLs, and out-of-scope observations,” which implies crawled free text/structured extraction is read back into the agent/LLM context.