agent-sandbox enables easy management of isolated, stateful, singleton workloads, ideal for use cases like AI agent runtimes and reinforcement learning (RL).
Agent Sandbox is a Kubernetes extension for managing isolated, stateful workloads that run as single long-lived containers with stable identities and persistent storage. It is intended for AI agent runtimes, reinforcement-learning workloads, and other applications that do not fit ordinary stateless deployments. Its catalogue skills and instructions support operating these sandbox workloads.
Latest release v1.0.7 · 15 Oct 2026
9 files for Claude Code, Codex, GitHub Copilot and OpenCode: bump-go-version, dev-rules, fix-flakes, k8s-api-conventions and 5 more — 5,623 tokens loaded in every session.
.github/copilot-instructions.md A 293 tok AGENTS.md A 5,325 tok CLAUDE.md A 5 tok .agents/skills/bump-go-version/SKILL.md A 25 tok .agents/skills/dev-rules/SKILL.md A 13 tok .agents/skills/fix-flakes/SKILL.md A 80 tok .agents/skills/k8s-api-conventions/SKILL.md A 20 tok .agents/skills/test-pyramid/SKILL.md A 104 tok .agents/skills/triage-issues/SKILL.md A 47 tok These files are kubernetes-sigs/agent-sandbox's own configuration — they tell Claude Code, Codex, GitHub Copilot and OpenCode how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.