Prompt Injection Mitigation

Implement confirmation guardrails for AI browser agents by asking them to confirm before sharing personal data and prefer vision-based DOM reading to mitigate hidden-text prompt injections.