browser-bridge

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The code intentionally implements a local browser bridge that removes CSP protections, exposes privileged extension APIs to arbitrary pages, reads and transmits cookies/tab data, allows remote arbitrary JavaScript execution in page contexts, and can enable/disable other extensions — all strong indicators of backdoor/exfiltration functionality.

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

• Third-party content exposure detected (high risk: 0.78). It navigates to/operates on runtime web pages (outsider-authored content) and then returns readable page-derived text/HTML fields (e.g., scan’s html and exec’s js_return/transients / diff), which are then available for the agent to include in its LLM context—via the Browser Bridge → Python → CLI JSON result path.