agent-integrity-monitor - AI Agent Integrity Monitoring
Monitor autonomous AI agents for integrity violations including prompt injection and behavioral deviation
Tags
Updated: 2026-03-23Capabilities
Typical Inputs
Typical Outputs
What this skill does
- detect prompt injection attempts
- detect instruction override
- detect goal drift
- detect behavioral deviation
- monitor agent sessions
- raise integrity violations
- classify violation types
- assess severity levels
- check trust model status
- identify repeated violations
- recommend remediation actions
Inputs
- SecurityFact
- USAP agent runtime
- production agent sessions
Outputs
- integrity violation report
- JSON violation metadata
- security incident escalation
Requirements
- admin invoke role
- admin approver role
- LLM provider access