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 agentmods add instructions/celestoai/smolvm/agents-mdgit clone --depth 1 https://github.com/CelestoAI/SmolVMWrote 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/celestoai/smolvm/agents-md)<a href="https://agentmods.dev/instructions/celestoai/smolvm/agents-md"><img src="https://agentmods.dev/badge/instructions/celestoai/smolvm/agents-md.svg" alt="Measured on agentmods" 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 | $0.01260 | $0.01260 |
| Opus 5 | $0.00630 | $0.00630 |
| Sonnet 5 | $0.00252 | $0.00252 |
| Haiku 4.5 | $0.00126 | $0.00126 |
Grade A, and why
SmolVM 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 4d ago.
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.
How it starts
The opening of the file, as written. The whole thing — 114 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SmolVM Context
SmolVM gives AI agents their own disposable computer. Each sandbox is a lightweight virtual machine that boots in seconds, runs any code or command you throw at it, and disappears when you're done — nothing touches the host.
🚀 Project Overview
SmolVM is specifically designed to provide a secure "sandbox" for AI agents to execute code, browse the web, or perform system-level tasks safely.
🧪 Development
Key Commands
- Testing:
pytest(runs the suite intests/) - Linting & Formatting:
uv run ruff check .oruv run ruff format .
Release checklist
- For guest-agent or published-image changes, build and smoke the new image release before tagging the SmolVM package release.
- Update
src/smolvm/images/published.pywith the newIMAGES_RELEASE_TAGand rootfs SHA pins before the PyPI tag is pushed. - Also update
src/smolvm/images/builder.py::_GUEST_AGENT_RELEASE_SHA256from thesmolvm-guest-agent-linux-<arch>.sha256release assets. This is a separate pin from the rootfs manifest. - Verify both paths:
uv run smolvm ...from a source checkout may build or use the local guest-agent binary, whilesmolvm ...fromuv tooluses the installed wheel and downloads the standalone guest-agent release binary. - Only tag the SmolVM package release after the image manifest, guest-agent binary SHA pins, image smoke, and focused tests are complete.
CLI design
- New CLI commands follow a NOUN-VERB structure:
smolvm <noun> <verb>, e.g.smolvm codex start, notsmolvm start codex. - The noun names the resource (a sandbox, a harness, a browser session); the
verb names the action on it (
start,stop,ssh). - This scales naturally as actions grow:
smolvm codex start, then latersmolvm codex logs,smolvm codex status, etc. - When adding a new harness or resource, register it as a top-level subcommand and put its actions underneath, instead of overloading a global verb.
- Documented exceptions: the top-level
smolvm pruneandsmolvm imagesaliases exist for muscle memory (imagesmirrorsdocker images); both delegate to their NOUN-VERB homes (image prune,image list). Do not add further top-level aliases without discussion. smolvm completion <shell>is an accepted top-level meta-command (likegit/ghcompletion): it acts on the CLI itself rather than a sandbox resource, so it sits outside the NOUN-VERB rule by design.
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.
- 4d ago First seen · 114 lines · 1,260 tokens per session scan A 94f684f422e8
SmolVM AGENTS.md is an instructions file published in the GitHub repository CelestoAI/SmolVM (866 stars, last pushed 7d ago), licensed Apache-2.0. It adds 1,260 tokens to every session, about $0.0063 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
Operator-Use CLAUDE.md
Instructions for Jeomon/Operator-Use, covering operator — claude.md, project overview, repository layout, quick navigation and key invariants.
browser-harness AGENTS.md
AGENTS.md instructions for browser-use/browser-harness, covering code priorities, commands, doctor — install/daemon/browser state, unit tests (no live browser) and security.
browser-harness CLAUDE.md
Claude Code instructions for browser-use/browser-harness, a project described as: Browser Harness | Self-healing harness that enables LLMs to complete any task.
arcbox CLAUDE.md
Claude Code instructions for arcboxlabs/arcbox, covering repository guidelines for coding agents, project overview, performance targets, platform priority and project structure.
arcbox copilot-instructions.md
Copilot instructions for arcboxlabs/arcbox, covering arcbox ai agent instructions, architecture overview, critical platform differences, development workflows and required once.
MultiGen CLAUDE.md
Instructions for lingyuanli/MultiGen, covering claude.md, architecture, development commands, full stack (docker) and api (local).