Skip to content

Security

This is the canonical home for all security documentation in Agent Governance Toolkit.

Start here

Topic Page
How threats are modeled across AGT's trust boundaries Threat Model
Multi-tenant isolation guarantees and operator checklist Tenant Isolation · Checklist
Trust score calibration methodology and thresholds Trust Score Calibration
Plugin and dependency scanning that runs on every PR Security Scanning
How AGT maps to the OWASP Agentic Top 10 OWASP ASI coverage
Dated security audit notes (additive contracts, sandbox extensions, etc.) Audits
Reporting a vulnerability Disclosure

Scope

This section covers the runtime security posture of AGT: how the policy engine, identity, sandbox, and audit subsystems defend against the threats documented in the threat model, and how operators verify that posture in production.

It does not cover:

  • Compliance framework mapping (NIST AI RMF, EU AI Act, SOC2, ISO 42001 et al.) — see the per-framework pages under docs/compliance/, e.g. NIST AI RMF alignment and SOC2 mapping.
  • Release security or supply-chain attestation for shipped artifacts — see the release notes under docs/releases/ and the SBOM workflow at .github/workflows/sbom.yml.

Reporting a vulnerability

If you believe you have found a security vulnerability in this repository, please follow the disclosure process in disclosure.md. Do not file a public GitHub issue.