Authoring Conventions - Superwire

Use snake_case names

schema task_summary { ... }
agent summarize_task { ... }
model fast from llm { ... }

Prefer intent-based model names

model fast from llm {
    id: "gpt-4.1-mini"
}

model smart from llm {
    id: "gpt-4.1"
}

The workflow can express why an agent uses a model without repeating provider-specific IDs everywhere.

Keep examples minimal

Do not introduce multiple models, tools, prompts, resources, or MCP servers until the page is specifically explaining those features.

Use regular comments and doc comments

// This is a regular comment.

schema release_note {
    /// Field description preserved for schema-aware tooling.
    title: string
}