GitHub's AI agents gain Docker Sandbox isolation in CI pipelines

GitHub Agentic Workflows now supports Docker Sandboxes (sbx) as an agent runtime environment, allowing AI coding assistants to safely execute commands within isolated microVMs while maintaining strict network and repository access controls. This integration, shipped in version 0.82.9 of the GitHub CLI extension, enables agents to perform tasks like running Java test suites with PostgreSQL via Testcontainers, fixing bugs, and creating draft pull requests - all within a disposable sandbox environment with sudo privileges. The setup requires no custom configuration and maintains GitHub Actions' existing CI infrastructure while adding containerized isolation boundaries. Workflow designers can fine-tune access by allowlisting specific network destinations and file paths, with pull requests automatically created in draft mode for human review.