autobrowse

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

• Third-party content exposure detected (high risk: 1.00). The skill's inner agent (scripts/evaluate.mjs) autonomously runs "browse open " and "browse snapshot"/"browse get text body" on URLs supplied in task.md or free-form user input (SKILL.md, evaluate.mjs), meaning it fetches and reads arbitrary public web pages and their accessibility/screenshot content which the agent then uses to choose clicks, fills, and next actions—so untrusted third‑party content can materially influence behavior.