Skip to main content
Agent skill

Cloud Image Bake

Bake, select, prove, and safely retire a Cloud Worker image with crabbox and config one-liners.

Get started

Get the complete skill folder, including its scripts and reference files.

When to use this skill

Use Cloud Image Bake when an AI agent needs a reusable SKILL.md workflow for this job: Bake, select, prove, and safely retire a Cloud Worker image with crabbox and config one-liners.

When not to use it

Skip Cloud Image Bake when the task is outside the coding category, or when a more specific skill in this directory already covers the same workflow with clearer triggers.

How to install

Where will you use it?

Install scope
Claude Code folder setup
mkdir -p ~/.claude/skills/cloud-image-bake
# Save SKILL.md to ~/.claude/skills/cloud-image-bake/SKILL.md
  1. Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/cloud-image-bake/SKILL.md. For a shared project, commit the folder at .claude/skills/cloud-image-bake/ instead.
  2. Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
  3. For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.

Full install guide for Claude, Cursor, and Codex →

Try a starter prompt

After installing the skill, paste this starter prompt into your agent to verify it triggers on a small task.

Starter prompt
Use the Cloud Image Bake skill for this task: Use Cloud Image Bake when this documented workflow matches the task. Confirm any required inputs or files before making changes.

What supports this recommendation?

68/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.

Documentation
69/100
Practical detail
68/100
Source evidence
63/100
Source signals
72/100
Model compatibility evidence
  • claudeuntested

    No model-specific signal or recorded compatibility test was found.

  • chatgptuntested

    No model-specific signal or recorded compatibility test was found.

  • geminiuntested

    No model-specific signal or recorded compatibility test was found.

  • copilotuntested

    No model-specific signal or recorded compatibility test was found.

  • llamauntested

    No model-specific signal or recorded compatibility test was found.

  • perplexityuntested

    No model-specific signal or recorded compatibility test was found.

  • mistraluntested

    No model-specific signal or recorded compatibility test was found.

  • grokuntested

    No model-specific signal or recorded compatibility test was found.

Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.

What this skill does

# Bake a Cloud Worker image

Never print or persist secret values; provider credentials stay in their stores. Never hand-edit config files on disk — profile changes go through `openclaw config`. Every run ends with the observable Prove result or an exact explanation of why it could not be proven. Snapshots are cheap; unmanaged snapshot sprawl is not. Never delete a provider image without hard operator confirmation.

## Gather

``` openclaw config get cloudWorkers --json crabbox config show --json crabbox doctor --provider <backend> --json crabbox checkpoint list --json ```

Record the current provider, class, image selection, setup command, and the id of the image being superseded. Confirm the requested tooling and a secret-free bake source.

## Mutate

Lease from the current profile, install and smoke-test the tooling:

``` crabbox warmup --provider <backend> --class <class> --keep --timing-json crabbox run --provider <backend> --id <lease> --no-sync -- bash -lc '<install commands> && <tool> --version' ```

Snapshot per backend:

- AWS: `crabbox checkpoint create --provider aws --id <lease> --mode native --strategy image --wait`, inspect it, then `crabbox image promote <ami-id>` with the matching scope. AWS image selection is owned by the promote catalog. - Hetzner: `hcloud image create --type snapshot --server <server-id> --description <name>`; there is no crabbox create/promote lifecycle for Hetzner yet, so record the snapshot id explicitly. - Firecracker: rebuild and republish the rootfs template through the host's template pipeline; do not snapshot a running microVM as a substitute.

Point the profile at the new selection only through validated config writes — confirm the exact key first, dry-run, then write (example for a backend whose settings carry an image field):

``` openclaw config schema --json | jq '.properties.cloudWorkers' openclaw config set cloudWorkers.profiles.<profile>.settings.<imageKey> "<image-id>" --dry-run openclaw config set cloudWorkers.profiles.<profile>.settings.<imageKey> "<image-id>" ```

The bundled crabbox profile currently has no `image` settings key — AWS selection lives in `crabbox image promote`; never invent a config field. Preserve the old image until proof passes.

## Repair

``` openclaw doctor --lint crabbox doctor --provider <backend> --json ```

`doctor --lint` can exit `1` for findings: read the report and continue the remaining checks. Ordinary `doctor` and `doctor --non-interactive` can write config/state; do not use them for diagnosis before approval. Apply `openclaw doctor --fix --non-interactive` only after explicit approval, then re-read the profile and provider inventory.

## Prove

Lease once from the new image and verify the baked tooling is present and fast:

``` crabbox warmup --provider <backend> --class <class> --timing-json crabbox run --provider <backend> --id <lease> --no-sync -- bash -lc '<tool> --version' crabbox stop --provider <backend> --id <lease> ```

Record warmup total and compare against the pre-bake timing. Then confirm the OpenClaw path end to end: dispatch one session to the profile from a client (Cloud destination) and verify the placement reaches active. If any step fails, roll back the image selection and report the exact blocker.

## Report

Report the profile, backend, new and previous image ids, tooling smoke result, timed warmup before/after, and rollback state. Only after successful proof, show the exact deletion target and get hard operator confirmation, then delete only that superseded snapshot (`crabbox image delete <id>` or `hcloud image delete <id>`) and verify it is gone. Without confirmation, leave it intact and report cleanup pending.

Further reference: https://docs.openclaw.ai/gateway/cloud-workers

Intended uses

  • Use Cloud Image Bake when this documented workflow matches the task.

Implementation notes

  • Review the source instructions and adapt inputs before running the workflow.

Try it, then tell us how it went

A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.