Bootstrap A2a Agent

Step 2: Confirm with User

Before taking any action, present the derived values and ask the user to confirm:

I'm about to create the following A2A agent repository:

- **Repository:** {organization}/{repository_name}
- **Template:** {organization}/acme-agent
- **Agent skill:** Derived from repository name (e.g., "travel-agent" → skill "travel")

Shall I proceed?

Do not continue until the user explicitly confirms.

Step 3: Create Repository from Template

Use the GitHub MCP server to create a repository from the template:

mcp_github_create_repository(
    name="{repository_name}",
    organization="{organization}",
    description="A2A agent: {description}",
    template_owner="{organization}",
    template_repo="acme-agent"
)

Step 4: Trigger Cookiecutter Workflow

Run the cookiecutter workflow to populate the repository:

mcp_github_run_workflow(
    owner="{organization}",
    repo="{repository_name}",
    workflow_id="Cookiecutter",
    ref="main"
)

Step 5: Provide Summary

Created A2A agent repository at https://github.com/{organization}/{repository_name}!

The repository has been populated with an A2A protocol agent. The structure includes:

- Express server with A2A protocol support
- Vercel AI SDK integration with configurable LLM provider
- MCP server support (stdio and HTTP transports)
- Docker and Compose configuration with watch mode
- Agent card, skills, and behavior defined in package.json
- Customizable model, tools, and MCP servers in src/agent/config.ts

Next steps:
1. Clone the repository
2. Update `package.json` with your agent's role, skills, and behavior
3. Configure your LLM provider and tools in `src/agent/config.ts`
4. Run `docker compose up` to start developing