upss-security-guard

Preventing prompt injection requires a runtime 6-gate security pipeline that intercepts and validates every prompt before LLM execution. This approach blocks malicious attempts, role confusion, and secret data leakage automatically during agent operations.

Runtime prompt validation applies security checks across agent prompts, tool inputs, and generated instructions before execution. It uses pluggable primitives like BasicSanitizer and InputValidator to enforce strict controls automatically within frameworks like OpenClaw.

Yes, you can enforce RBAC and audit logging for LLM prompts to maintain least-privilege access and compliance. The security pipeline logs every prompt evaluation, providing a clear audit trail of all runtime validations and policy enforcement actions.

OpenClaw supports automatic security policy enforcement through a dedicated plugin that protects both agent prompts and tool inputs. This integration applies the 6-gate validation pipeline seamlessly across your existing agent framework operations.

Yes, you need sqlite3, openssl, and sha256sum as dependencies to run these prompt security checks. These underlying tools support the local audit logging, cryptographic verification, and data integrity required by the runtime security pipeline.

Lightweight runtime prompt auditing focuses on intercepting known high-risk patterns like prompt injection rather than deep semantic analysis. It relies on pluggable primitives like LightweightAuditor and SimpleRBAC to enforce strict policy controls without heavy computational overhead.