cyberstars
cyberstars / tools / agent-governance-toolkit

Agent Governance Toolkit

Policy enforcement, sandboxing and audit for autonomous AI agents

Voir sur GitHub
PythonMIT

Microsoft framework that intercepts AI agent tool calls at the application middleware layer, allowing or denying actions deterministically against YAML/OPA/Cedar policies. Adds identity verification, execution sandboxing and tamper-evident audit logging, with coverage mapped to the OWASP Agentic Top 10 and integrations for OpenAI Agents SDK, CrewAI and LangGraph.