A
licenseNot graded
qualityB
maintenanceEnables agents to screen inputs, tool arguments, and retrieved documents for prompt injection, jailbreak, privilege escalation, tool coercion, exfiltration, and PII, returning structured allow/refuse/sanitize verdicts with masked redactions. Runs fully offline and deterministically with no model calls, network access, or third-party dependencies, exposing screening, redaction, and moderation tools over the Streamable HTTP transport.
MIT