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.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add kubernetes-sigs/agent-sandbox --skill skillgit clone --depth 1 https://github.com/kubernetes-sigs/agent-sandboxWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/kubernetes-sigs/agent-sandbox/skill)<a href="https://agentmods.dev/skills/kubernetes-sigs/agent-sandbox/skill"><img src="https://agentmods.dev/badge/skills/kubernetes-sigs/agent-sandbox/skill/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/kubernetes-sigs/agent-sandbox/skill"><img src="https://agentmods.dev/badge/skills/kubernetes-sigs/agent-sandbox/skill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 8 Skill grants unrestricted tool access without appropriate constraints. An agent with unfettered tool access can perform arbitrary actions including file modification, network requests, and code execution.Fix: Restrict tool access to only the tools required for the skill's stated purpose. Use an explicit allowlist rather than granting blanket access.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00046 | $0.01422 |
| Opus 5.5 | $0.00018 | $0.00569 |
| Sonnet 5.5 | $0.00009 | $0.00284 |
| Haiku 4.5 | $0.00005 | $0.00142 |
Grade B, and why
k8s-agent-sandbox-mcp scanned grade B with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Unrestricted tool accessmediumExcessive agency
A wildcard tool grant or "run any command" leaves no least-privilege boundary at all.
> **⚠️ SECURITY WARNING:** This MCP server does not have built-in authentication. It exposes tools that can create sandboxes, execute arbitrary commands, and read or write files using the permissions of its Kubernetes se Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
3. Tool Call: `execute_command(sandbox_claim_name="sbx-12345", namespace="default", command="python -c 'import requests; print(requests.get(\"https://example.com\").status_code)'")` How it starts
The opening of the file, as written. The whole thing — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Kubernetes Agent Sandbox Manager (MCP)
⚠️ SECURITY WARNING: This MCP server does not have built-in authentication. It exposes tools that can create sandboxes, execute arbitrary commands, and read or write files using the permissions of its Kubernetes service account. Before proceeding, ensure that the server is strictly isolated on a private network or secured behind an authenticating proxy.
Use this skill when a task requires running code, executing untrusted scripts, or performing heavy parallel workloads in an isolated Kubernetes environment. This skill connects to the official kubernetes-sigs/agent-sandbox MCP server. It can be configured using mcp-config.json in this skill's directory.
Architecture & State
Unlike basic shell execution, this sandbox is stateful.
When you create a sandbox, a persistent Kubernetes Pod is provisioned and identified by a sandbox_claim_name. You can run multiple sequential commands against the same sandbox_claim_name (e.g., install a package, then run a script). You must retain the sandbox_claim_name value in context for the lifetime of the task.
Available MCP Tools
-
create_sandbox- Arguments:
warmpool(string),namespace(string),sandbox_ready_timeout(int, optional),labels(dict[string, string], optional),shutdown_after_seconds(int, optional, defaults to 300 — the sandbox self-deletes after 5 minutes unless raised),pod_labels(dict[string, string], optional),pod_annotations(dict[string, string], optional). - Returns: JSON object with
sandbox_claim_name. - Purpose: Create a new sandbox.
- Arguments:
-
execute_command- Arguments:
sandbox_claim_name(string),namespace(string),command(string — shell command or python inline script),timeout(int — seconds before the command times out, optional),max_output_chars(int, optional, default 30000, minimum 128, maximum 300000). - Returns: JSON object containing
stdout,stderr,exit_code,stdout_total_chars,stderr_total_chars,stdout_truncated, andstderr_truncated, or JSON with anerrorfield. - Output budget: Limits the combined stdout/stderr Unicode code points, including omission markers. Both streams initially share the budget equally, with unused capacity reassigned to the other stream. Oversized streams retain equal-sized beginning and end excerpts (at most one extra character at the end), separated by
[... N characters omitted ...]. The*_total_charsfields report SDK output lengths before this tool truncates them;*_truncatedindicates whether the returned output is partial. - Purpose: Executes commands inside the provisioned sandbox.
- Arguments:
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday Changed · +1 lines 37f79a45e5a6
- 1mo ago First seen · 72 lines · 46 tokens per session scan B 14f330980995
k8s-agent-sandbox-mcp is a skill published in the GitHub repository kubernetes-sigs/agent-sandbox (4,346 stars, last pushed today), licensed Apache-2.0. It adds 46 tokens to every session and 1,422 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it B with 2 findings (unrestricted tool access, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-05.
Other skills, from other repositories
04-security
A security and operations guide for deploying Dify, an open-source platform for building AI applications, in production.
troubleshoot-sandbox
Troubleshoot OpenSandbox issues by running diagnostics (logs, inspect, events, summary) via CLI or HTTP API to diagnose sandbox failures like OOM, crash, image pull errors, network problems, etc.
mirrord-kafka
Helps DevOps engineers configure mirrord Operator's Kafka queue splitting feature end-to-end. Generates the MirrordSplitConfig and MirrordPropertyList Kubernetes CRD YAMLs (the current resources; MirrordKafkaTopicsConsumer + MirrordKafkaClientConfig are deprecated but still supported), the matching mirrord.json…
mirrord-propagation-discovery
Scans a live Kubernetes environment with mirrord dump to find an identifier mirrord can filter traffic on: a header, a baggage or tracestate member, a tenant or user id, a trace id (W3C, B3, X-Ray, Google, Jaeger), a JWT claim, a cookie, a path segment, or a JSON body field that is present on every hop of a request…
langbot-deploy
Deploy and configure a LangBot instance — Docker / Docker Compose, Kubernetes, the config.yaml model, the Box sandbox runtime, the plugin runtime, and the global API key. Use when installing, deploying, upgrading, or configuring LangBot in production or self-hosted environments. Triggers on "deploy langbot", "langbot…
timoni
Use when deploying applications to Kubernetes with Timoni. Covers installing and upgrading module instances from OCI registries, composing multi-app deployments with bundles, injecting values from clusters or CI with runtimes, targeting multiple clusters, and authoring, testing, signing and publishing modules with CUE.