Last released Jul 1, 2026
A quarantine line for the data your LLM agent ingests: scan untrusted text and outbound actions for prompt injection and exfiltration.
Supported by
