An OpenHands skill that lets an OpenHands agent run work in a Runzivo sandbox: a Linux VM in your Runzivo workspace. The agent creates the sandbox, connects over SSH, runs commands, reports status, cost and expiry, stops and starts it, and deletes it after you confirm.
The skill is Markdown only. It calls the rzv CLI that you already use; it has no code of its own, stores no tokens and contains no AI provider keys.
rzv0.26.0 or newer, signed in withrzv login. Install it withcurl -fsSL https://runzivo.dev/downloads/rzv/install.sh | sh.- OpenSSH on macOS or Linux for the
<sandbox-id>.rzvalias. Without it the agent usesrzv sandboxes exec. - An OpenHands version that loads Agent Skills from
.agents/skills/(the CLI 1.16.0 and Agent Canvas do).
For every project (user skill):
git clone --depth 1 --branch v0.1.0 https://github.com/Runzivo/rzv-openhands.git
mkdir -p ~/.agents/skills && cp -R rzv-openhands/skills/runzivo ~/.agents/skills/For one project, copy skills/runzivo to <project>/.agents/skills/runzivo instead. The skill loads when your message mentions Runzivo, or when the agent decides it needs it.
Ask OpenHands, for example:
- "Create a Runzivo sandbox and run the test suite there."
- "What is the status of my Runzivo sandbox, and how much has it cost?"
- "Stop the Runzivo sandbox." / "Delete the Runzivo sandbox."
Sandboxes the skill creates are named openhands-<slug>. More in docs/usage.md.
bin/rzv-openhands-backend up starts an OpenHands Agent Server in a Runzivo sandbox and tunnels it to http://127.0.0.1:8000 over SSH, for Agent Canvas or the SDK. See docs/backend.md.
| Page | What it covers |
|---|---|
| Usage | What the agent does, persistence, billing |
| Backend | The Agent Server in a sandbox, over an SSH tunnel |
| Security | Keys, sign-in and what the skill never does |
| Testing | Checks and the live end-to-end test |
| Releasing | Tags and the OpenHands extensions registry |
MIT. rzv and the Runzivo service have their own licences and terms.