# Install and use the Rover skill

Use this workflow to build a modular vehicle experiment with an AI coding agent,
or adapt the existing Rover Foundry assets. The skill contains instructions and
references; models and media are separate downloads.

## 1. Install Scenario's skills

With Node.js and npm available, run this in your project directory:

```sh
npx skills add scenario-labs/skills --skill "*"
```

Select your agent and installation scope when prompted. Installing the collection
includes the core `scenario` skill and the image, 3D, texture, skybox, audio and
model-specific skills used throughout this workflow. If they are already installed,
check that your agent can discover them. Follow the
[official installation instructions](https://github.com/scenario-labs/skills#install)
for supported agents and updates.

## 2. Connect Scenario MCP

Add `https://mcp.scenario.com/mcp` as a Streamable HTTP MCP server in your agent
and sign in with your Scenario account through OAuth. Skills provide guidance;
the MCP connection provides generation tools. The install command above does not
configure that connection.

Follow the [Scenario connection guide](https://mcp.scenario.com/docs?referral=experiments-rover-foundry&utm_source=scenario.com&utm_medium=experiment&utm_campaign=rover-foundry&utm_content=scenario-mcp)
for your client. Ask your agent to use the `scenario` skill and list your teams to
verify access without generating anything. Choose your own team/project and agree
on a generation budget before creating new assets. Never paste credentials into
a prompt. Existing public asset downloads need no Scenario account.

## 3. Install the Rover skill

Extract the skill ZIP. From your project directory, run the command below,
replacing `/path/to/extracted` with the actual extraction directory:

```sh
npx skills add /path/to/extracted/scenario-modular-vehicle-experiment
```

Choose the same agent and project scope as above. Alternatively, copy the whole
folder into your agent's documented skill directory. Keep `SKILL.md`,
`references/` and `agents/` together. Ask your agent to confirm it can discover
`scenario-modular-vehicle-experiment`; refresh the agent session if needed.

The complete asset kit includes the same folder inside `Rover Foundry/`. Run the
local install command with the path to that folder when using the complete kit.
Quote the path if it contains spaces.

## 4. Choose a starting point

For a new experiment:

> Use $scenario-modular-vehicle-experiment to plan a four-wheel desert exploration
> vehicle configurator. Use Scenario's installed skills for generation. Start by
> checking my tools and asking for the project and budget; show the proposed
> equipment categories before generating assets.

For the downloaded Rover assets:

> Use $scenario-modular-vehicle-experiment to inspect the extracted Rover Foundry
> kit at [my folder]. Explain its Blender hierarchy and propose how to adapt it
> into a four-wheel vehicle. Start from the packed native and do not generate
> new assets yet.

Replace `[my folder]` with your actual extraction path. In clients without `$`
invocation, ask the agent to use the skill by name.

## Tools and files for each task

| Task                                        | What you need                                                                                                                     |
| ------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------- |
| Explore or download Rover Foundry           | A browser; no login or local tool installation                                                                                    |
| Generate new assets                         | An agent with the Scenario skills, an authenticated Scenario MCP connection, an authorized project and available credits          |
| Edit the supplied model or run its exporter | Blender compatible with the native file (the kit was authored in Blender 5.2.1), plus a terminal for the included Python exporter |
| Automate Blender through an agent           | A connected Blender integration exposing the required operations; the original used Scenario Blender Studio, installed separately |
| Build a new browser configurator            | A coding agent, local HTTP preview and the development tools chosen for that project                                              |

Check the installed Blender integration's capabilities before relying on it.
Scenario's skills and remote MCP do not install or control Blender by themselves.
If no integration is available, choose manual Blender steps or a local script
workflow with the user and describe the actual method used.

Get the [complete asset kit](https://www.scenario.com/experiments/rover-foundry/downloads/rover-foundry-complete-kit.zip),
[editable Blender model](https://www.scenario.com/experiments/rover-foundry/downloads/rover-foundry.blend),
[recipe](https://www.scenario.com/experiments/rover-foundry/downloads/rover-foundry-recipe.md)
and [exact requests](https://www.scenario.com/experiments/rover-foundry/downloads/rover-foundry-prompts.json).
The complete kit includes assets, the exporter, recipe, PDF and this skill. It
does not include runnable HTML/JavaScript, a website development checkout or the
creator's private production workspace. Use the
[hosted configurator](https://www.scenario.com/experiments/rover-foundry/) to explore
the finished experience; use this workflow to author your own application.

Historical request IDs document the original work. Replaying them requires access
to the original private project; for new generations, use your own project and
upload local reference assets to obtain your own IDs. Discover currently available
models rather than assuming historical IDs are available to your account.