Skip to main content

​
0.6.0 (2026-08-18)

  • Removes the unused DockerSandbox API and the direct Testcontainers/docker-java dependency.
  • Keeps local process execution, the Sandbox abstraction, LocalSandbox, Agent Judge grading, the agents module, and its real Agent Client-backed LLM judge.
  • Documents that workspaces are not security boundaries and recommends externally managed isolation for untrusted benchmark definitions or agents.
  • Aligns the dependency floors with Jackson 2.21.6, Jackson 3.1.6, Agent Judge 0.14.0, and Agent Client 0.26.0.
  • Publishes agent-bench-core and agent-bench-agents 0.6.0, plus the aggregate CycloneDX 1.6 SBOM.

​
0.5.0 (2026-08-17)

  • Added the Spring Boot upgrade benchmark candidate and retained its owner-operated external baseline-aware grader distinction.
  • Published the root aggregate CycloneDX SBOM and aligned Agent Judge and Agent Client dependencies.
  • Released two Maven modules: agent-bench-core and agent-bench-agents.

​
0.4.0 (2026-06-06)

  • Aligned Jackson 2 and Agent Client dependencies.

​
0.3.0 (2026-05-19)

  • Migrated to the markpollack organization, io.github.markpollack Maven coordinates, and the Business Source License 1.1 for new development.
  • Added Central metadata and documentation links.

​
0.2.1 (2026-03-31)

  • Initial release.