@wireweave/agent-prompts

Canonical grammar prompts for LLM agents that generate Wireweave DSL. The package is the single source of truth so consumers (agent-harness, api-server, future plugins) stay in sync with the multi-page canvas model in @wireweave/core 3.x.

Install

pnpm add @wireweave/agent-prompts

Use

import { buildGrammarPrompt, buildCompactGrammarPrompt } from '@wireweave/agent-prompts'

// Full prompt — generation phase.
const system = buildGrammarPrompt()

// Compact prompt — analyze / plan phases (saves ~800 tokens).
const compact = buildCompactGrammarPrompt()

Both functions return a single Markdown-flavoured string that can be injected verbatim into an LLM system message. No runtime dependencies.

What's covered

• Single-page wireframes (legacy).
• Multi-page canvas — multiple top-level page declarations, at(x, y) positioning, viewport="WxH".
• Renderer surface from @wireweave/core: render(doc) / renderCanvas(doc) / renderPage(page).
• Layout / content / form / nav / data / feedback / overlay / utility component contracts.
• Mapping common UI concepts to DSL components.

Why a separate package

Before this package the DSL grammar guide was duplicated in api-server/src/services/guide.ts, api-server/src/agent/prompts.ts, and agent-harness/wireweave/src/guide.ts. The agent-harness copy drifted to a pre-3.0 single-page-only definition, which silently broke multi-screen wireframe generation. Pulling the guide into its own package gives a single versioned export every downstream consumer can pin.

License

MIT