openclaw-agent-skills
Shared skills for coding agents that work on OpenClaw projects. This repo is the public canonical source for common workflows such as review closeout and remote validation.
Crabbox runs your repository's commands on remote machines, existing SSH hosts, and local containers. Keep editing locally; send your working tree to a box, stream the output, and get the command's exit code back.
$ npx -y skills add openclaw/crabbox --agent claude-code
Repo: openclaw/crabbox
What's inside
Crabbox runs your repository's commands on remote machines, existing SSH hosts, and local containers. Keep editing locally; send your working tree to a box, stream the output, and get the command's exit code back.
Quick start · Commands · Install · Providers · Documentation · Security
Inside a Git repository you trust, with Docker or Podman running:
crabbox run \
--provider docker \
-- uname -a
One command creates a Linux container, syncs your checkout, runs uname -a,
streams the result, and removes the container. No cloud account or login needed.
For remote compute, choose a provider or your team's coordinator.
Homebrew installs the complete release distribution:
brew install openclaw/tap/crabbox
crabbox --version
For macOS, Linux, and Windows, you can also download a release archive. Windows users should follow the Windows installation guide.
When a release includes a crabbox-runtime directory, keep it beside the real
CLI executable. New filesystem-capable packs contain amd64 and arm64 companions
for Linux, macOS, and Windows. Linux companions additionally provide native
managed execution for Linux and WSL2; copying only the CLI omits those bundled capabilities.
Do not mix a runtime pack with a differently built controller.
An official release with a missing or incomplete pack is a broken installation:
reinstall its matching archive or Homebrew package. It does not silently compile
replacement companions from source. CLI-only source builds retain their existing
shell-backed supervisor route.
Local prerequisites for the SSH workflow: git, ssh, ssh-keygen,
rsync, and curl. The local quick start also needs a running Docker or
Podman engine.
go install github.com/openclaw/crabbox/cmd/crabbox@v0.44.0
This install channel is supported starting with v0.44.0. Use an explicit release
version; do not use @latest while older, incompatible releases remain visible.
The module requires Go 1.26 and prefers go1.26.5; use Go 1.26.5 or newer, or
leave automatic toolchain selection enabled.
go install compiles only the CLI. It omits companion executables and assets,
including crabbox-apple-vm-helper and the native runtime pack, and is not the
signed/notarized prebuilt distribution. Use Homebrew or a release archive for complete platform capabilities,
especially Apple VM support on Apple Silicon.
Without an installed pack, source-built CLIs can compile the dependency-free
filesystem helper with a local Go 1.26-or-newer compiler. This does not compile
the managed-command supervisor or produce a signed release installation.
Inside a Git repository you trust, with Docker or Podman running:
crabbox doctor --provider local-container
crabbox run --provider local-container -- uname -a
You should see the container's Linux kernel information. Crabbox returns the command's exit code and cleans up the one-shot container. First startup includes an image pull and bootstrap, so allow extra time.
The default box is bare. It includes Crabbox's sync/run prerequisites, but
not your project's Node, Go, Python packages, or other build dependencies.
Prepare those before replacing uname -a with your test command.
For a Node.js repository with a package-lock.json and an npm test script,
select an image with Node installed and install the dependencies on the box:
crabbox run --provider local-container \
--local-container-image node:22-bookworm \
--shell 'npm ci && npm test'
Choose a Node version compatible with your project. Crabbox does not install project runtimes for you: supply them with a prepared image, the repository's own setup scripts, a devcontainer, Nix, or mise/asdf. Crabbox can also reuse supported setup steps from GitHub Actions; full Actions semantics require the documented runner path.
crabbox warmup --provider local-container --slug dev-box
crabbox run --provider local-container --id dev-box -- uname -a
# Edit locally, then run again on the same box.
crabbox run --provider local-container --id dev-box -- uname -a
# Open a shell, or release the box when finished.
crabbox ssh --provider local-container --id dev-box
crabbox stop --provider local-container dev-box
Replace uname -a with your test command once the box is prepared. warmup
creates a reusable lease; prewarm additionally performs Actions hydration.
A lease has both a stable cbx_... ID and a friendly slug; either works with
--id.
For a complete repository setup, see Getting started and Local Container.
Use the URL supplied by your team; the address below is a placeholder:
crabbox login --url https://broker.example.com
crabbox doctor
crabbox run -- pnpm test
This assumes the repository's provider and remote runtime/dependency setup are configured. For your own cloud account, follow a direct-provider setup guide and skip login. Cloud runs use your infrastructure and may incur charges.
| Command | What it does |
|---|---|
crabbox doctor | Check local prerequisites, configuration, and provider reachability. |
crabbox run -- <cmd> | One-shot: get a box, sync, run, stream output, release. |
crabbox run --shell '<script>' | The same, for a multi-step shell command. |
crabbox warmup / crabbox prewarm | Create a reusable box; prewarm also runs Actions hydration. |
crabbox run --id <box> -- <cmd> | Reuse a warm box, syncing only what changed. |
crabbox ssh --id <box> | Open an interactive shell on the box. |
crabbox job run <name> | Run a named workflow defined in .crabbox.yaml. |
crabbox list / crabbox stop <box> | See active boxes, and release one when finished. |
Pass long commands as a file with --script <file> instead of a large quoted
string. Every command has a page under Commands, and
the CLI reference lists all flags and environment variables.
| Situation | Try this |
|---|---|
| The run never reaches the box | crabbox doctor --provider <name> for prerequisites and reachability. |
| You need to inspect the failure | Add --keep-on-failure, then crabbox ssh --id <box> into the exact box that failed. |
| A warm box behaves as if stale | Add --full-resync to reset the remote workdir before syncing. |
| Output is binary or terminal-hostile | --capture-stdout <path>, and --capture-stderr <path>. |
| You need a file the run produced | --download remote=local, repeatable for several files. |
Failed SSH-backed and Blacksmith delegated runs save local bundles to
.crabbox/captures/*.tar.gz by default, falling back to the Crabbox user state
directory when the project destination is unwritable. Follow the reported
failure-bundle local=… path; see local capture storage.
Crabbox does not scrub bundles, captured output, or artifacts; review them before
sharing.
Troubleshooting · History and logs · Observability
Your working tree
|
v
Select a box -> Sync changes -> Run command -> Stream output
^ |
| v
+------ Reuse the box, or release it
The CLI runs on your machine. For SSH providers, it sends files and runs commands directly on the runner over SSH and rsync. Delegated providers use their own execution transports.
An optional coordinator handles provider credentials, leases, budgets, history, and cleanup for supported managed providers. It is separate from the normal CLI-to-runner data path. Local containers and existing hosts work without a coordinator.
| Path | Use it when |
|---|---|
| No coordinator | You want local containers, existing hosts, or direct providers with local credentials and state. |
| Cloudflare Workers + Durable Object | You want the established managed coordinator deployment. |
| Node.js + PostgreSQL | You want to host the coordinator in containers, on a VM, or in Kubernetes. Complete the deployment proof before production cutover. |
State does not automatically migrate between coordinator runtimes. For deployment requirements and the separate SSM-only private AWS workspace path, see Infrastructure and Private AWS Workspaces.
Read How Crabbox works for the full lifecycle, Architecture for internals, and Vision for scope and non-goals.
Start with the environment you already have:
| Your starting point | Provider guides |
|---|---|
| Docker or Podman on your laptop | Local Container |
| An existing Linux, macOS, or Windows machine | Static SSH |
| A cloud account | AWS, Azure, Google Cloud, Hetzner |
| Your own virtualization infrastructure | Proxmox, Incus, Firecracker |
| Apple Silicon | Apple VM, Apple Container |
| A managed sandbox service | Daytona, E2B, Modal |
Showing a partial view of a very large repo.
Shared skills for coding agents that work on OpenClaw projects. This repo is the public canonical source for common workflows such as review closeout and remote validation.
FAQ
crabbox is a Claude Code plugin with 2 hand-picked skills for testing work, indexed on Flowy. Install it with the command on its page. It includes crabbox-quickstart, crabbox. Its skills do not fire on their own yet. Request auto-invocation to have Flowy route them as you prompt. Free and open source.
Is this plugin yours?
Claim it with GitHubSubmit a pluginPromote it