ArcBox is a Rust runtime for running Docker-compatible containers, disposable microVM sandboxes, and full Linux or macOS virtual machines on macOS. It is used to run AI agents, untrusted code, containers, and isolated machines locally.
Borrowing it
Nothing to install: this file belongs to arcboxlabs/arcbox. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/arcboxlabs/arcbox/master/AGENTS.mdgit clone --depth 1 https://github.com/arcboxlabs/arcboxWrote 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/instructions/arcboxlabs/arcbox/agents-md)<a href="https://agentmods.dev/instructions/arcboxlabs/arcbox/agents-md"><img src="https://agentmods.dev/badge/instructions/arcboxlabs/arcbox/agents-md/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/instructions/arcboxlabs/arcbox/agents-md"><img src="https://agentmods.dev/badge/instructions/arcboxlabs/arcbox/agents-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.04493 | $0.04493 |
| Opus 5.5 | $0.01797 | $0.01797 |
| Sonnet 5.5 | $0.00899 | $0.00899 |
| Haiku 4.5 | $0.00449 | $0.00449 |
Grade A, and why
arcbox AGENTS.md scanned grade A with 0 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
This is a copy
100% identical to skills AGENTS.md — 226 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 225 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repository Guidelines for Coding Agents
Project Overview
ArcBox is a macOS container and virtual machine runtime in Rust. Optimize local compatibility and performance against OrbStack.
Keep local sandbox SDKs and E2B compatibility active. Keep SDK publishing, Apple signing and notarization, and R2/CDN distribution active.
ArcBox Runners (runner/, ADR 0006) is an opt-in self-hosted GitHub Actions runner agent. Its control plane lives in arcboxlabs/platform. Local features must never require a Platform account.
The project is in alpha. Breaking changes (internal or user-facing) are acceptable — always prioritize consistency, coherence, and long-term maintainability, even if that means large-scale rewrites.
Performance Targets
| Metric | Target | OrbStack |
|---|---|---|
| Cold boot | <1.5s | ~2s |
| Warm boot | <500ms | <1s |
| Idle memory | <150MB¹ | ~200MB |
| Idle CPU | <0.05% | <0.1% |
| File I/O | ≥100% of the best competitor guest² | 75-95% (vs native, their claim) |
| Network throughput | >50 Gbps | ~45 Gbps |
¹ Host cost is the high-water mark of guest-touched pages, not the
configured memory_mb: a fresh idle 16 GB VZ VM measured ~718MB of
host phys_footprint (2026-08-01). On VZ that mark is a one-way
ratchet — guest RAM lives in Apple's XPC process, out of the daemon's
reach — so the guest freeing 3GB leaves the host paying for it forever.
On HV the guest's free page reporting drains it back (measured
2026-09-25 on a 16 GB VM: 1.9 GB → 5.0 GB while the guest held 3 GiB
of tmpfs → 1.95 GB ten seconds after it freed it). The mechanism, and
the madvise dead ends, are in virt/arcbox-vmm/AGENTS.md
"Releasing guest RAM" and engine/arcbox-engine/src/vm_lifecycle/balloon/mod.rs.
² Same-context comparison (container vs container, same-day pairing).
Cache-hot native is not a valid denominator for FUSE metadata —
methodology and current numbers in docs/benchmarks/virtiofs.md.
Platform Priority
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 · +35 tokens per session d1add545cd23
- 6d ago Changed · +1 lines · +100 tokens per session b2f253facd65
- 11d ago Changed · +2 lines · +294 tokens per session c73c8e38d935
- 1mo ago Changed · +4 lines · -189 tokens per session 4f38e128a5c2
- 1mo ago Changed · +65 tokens per session eb672dc818ef
- 1mo ago Changed · +1 lines · +257 tokens per session 30c08d3a5b98
- 1mo ago Changed · +18 lines · +286 tokens per session 83fe0bc151e8
- 1mo ago Changed · +1 lines · +104 tokens per session 43075e6f86a9
- 1mo ago Changed · +196 lines · +3,537 tokens per session 94549f3949a5
- 2mo ago First seen · 1 lines · 4 tokens per session scan A 6ebdb617a810
arcbox AGENTS.md is an instructions file published in the GitHub repository arcboxlabs/arcbox (4,251 stars, last pushed yesterday), licensed Apache-2.0. It adds 4,493 tokens to every session, about $0.0180 per session on Opus 5.5. A static security scan graded it A with 0 findings. It is 100% identical to skills AGENTS.md, differing in 226 lines, and is treated as a copy.
arcbox AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| arcbox AGENTS.mdarcboxlabs/arcbox | Instructions file | 4.3k | yesterday | Apache-2.0 | A | 4,493 | Codex, OpenCode |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | today | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | today | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | today | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 274k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |