Skip to content

AgentGuard v0.2.0 - Production-Grade AI Agent Authorization & Governance

Choose a tag to compare

View all tags
@Amaljeevs Amaljeevs released this 12 Oct 12:49
· 17 commits to main since this release
v0.2.0
d60b6aa

AgentGuard 0.2.0

Key Highlights

  • Verified Agent Identity: Never trust agentId/roles from payload bodies. Added AgentIdentityResolver, JwtClaimsAgentIdentityResolver supporting Keycloak and Entra ID claims, and SpiffeAgentIdentityResolver for workload SVIDs.
  • Action-Bound Human Approval: New �gentguard-approval module binding human approvals to cryptographic SHA-256 request hashes, with full APPROVAL_REQUIRED -> APPROVED / DENIED / EXPIRED -> CONSUMED lifecycle and single-use execution guarantees.
  • Secure Delegation: Enforce child permissions ? parent permissions, ancestor identity lookups, cycle detection, maximum delegation depth, and validity bounds.
  • Policy Engine Hardening: Deterministic policy evaluation, policy revision hashes, risk levels, rule IDs, and trace IDs.
  • Spring AI Integration: �gentguard-spring-ai module providing GuardedToolCallback to guard LLM tool calls.
  • Language-Neutral Authorization API: �gentguard-spring-web adapter exposing POST /v1/authorize with auto-configuration.
  • Sanitized Audit Evidence: Parameter sanitization masking secrets, recording hashes and execution provenance.
  • Comprehensive Adversarial Suite: Verified against privilege escalation, identity forging, approval tampering, replay attacks, and expired credentials.
Assets 2
Loading