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 skills/celestoai/smolvm/cli-docs-guidelinesnpx skills add CelestoAI/SmolVM --skill cli-docs-guidelinesgit clone --depth 1 https://github.com/CelestoAI/SmolVMWhat 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 | $0.00050 | $0.00751 |
| Opus 5 | $0.00025 | $0.00376 |
| Sonnet 5 | $0.00010 | $0.00150 |
| Haiku 4.5 | $0.00005 | $0.00075 |
Grade A, and why
cli-docs-guidelines 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 3d 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 — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLI Documentation Guidelines
Review or write CLI docs following these principles. CLI docs serve two audiences: newcomers running their first command, and experienced users scanning for flags.
Core Principles
All README Guidelines apply here. In addition:
1. Logical command ordering
Commands must appear in the order a user would run them. A command should never reference output or state from a command that appears later in the docs.
Wrong — stop appears before the user knows how to list:
smolvm stop <vm_id>
smolvm list
Right — create, inspect, then destroy:
smolvm create --name my-sandbox
smolvm list
smolvm stop my-sandbox
2. Introduce a concept before its flags
Show the base command before showing any flags or subcommands. Each flag is a new concept — don't introduce two flags in the same example unless they always go together.
Wrong — --os and --name are both new:
smolvm create --os debian --name my-debian-sandbox
Right — --name first, then a separate example for --os:
# Create a sandbox with a name
smolvm create --name my-sandbox
# Use a different OS image
smolvm create --os debian --name my-debian-sandbox
3. Show expected output after commands that produce it
When a command prints a value the user needs (an ID, a URL, a status), show it. The reader should never have to run the command to find out what it returns.
smolvm browser start --live
# Session: sess_a1b2c3
# Live view: http://localhost:6080
4. One topic per section
Don't mix sandbox lifecycle commands with browser session commands in the same section. Each distinct workflow gets its own heading.
5. Flags reference comes after prose explanation
Never lead with a flags table. Explain what the command does in plain language first, then list flags for readers who want to go deeper.
Review Checklist
- Commands appear in the order a user would run them
- Every placeholder (
<vm_id>,<session_id>) is introduced by a prior command or clearly labelled as "output from the previous step" - Each code block introduces at most one new flag or subcommand
- Commands that print useful output show that output as a comment
- Conceptually distinct workflows (e.g. sandbox vs. browser) are in separate sections
- Flags/options table, if present, appears after the prose description
- No jargon (SSH, TAP device, CIDR, firecracker, QEMU) without a plain-language explanation on first use
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.
- 3d ago First seen · 87 lines · 50 tokens per session scan A 7a3ec3f82069
cli-docs-guidelines is a skill published in the GitHub repository CelestoAI/SmolVM (866 stars, last pushed 6d ago), licensed Apache-2.0. It adds 50 tokens to every session and 751 once invoked, about $0.0003 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 skills, from other repositories
ai-deploy
Deploy a generated project (one process or several) to agent-sandboxes with scale-to-zero: idle sandboxes pause and auto-resume (process restarted) on the next HTTP request. Use when asked to deploy, host, or preview generated code through the sandbox platform.
e2b-sandbox
Create, manage, and use E2B sandboxes — run commands, manage files, use git, persist state, and configure networking. Use when building agent workflows that need isolated execution environments.
e2b-code-interpreter
Execute code in E2B sandboxes and integrate with LLMs for tool calling. Use when building AI agents that need to run Python/JS code, analyze data, generate charts, or use LLM function calling with E2B.
e2b
Core E2B SDK knowledge — correct imports, namespacing, terminology, and API patterns. Activates automatically when working with E2B sandboxes, templates, or code execution.
workflow-builder
Generates Python workflow files using the Operator workflow DSL. Knows all available patterns (classify-and-act, fan-out-and-synthesize, adversarial-verify, generate-and-filter, tournament, loop-until-done) and produces ready-to-run .py files saved to the workflows directory.
skill-creator
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.