Agentic Intent

Overview: SEO for LLMs

The same way websites use <meta> tags so search crawlers understand their content, AgnosticUI structures playbook recipes so AI agents understand your UI intent.

When a user asks their AI assistant "build me a dashboard," the assistant knows to reach for ag-card for metric tiles, ag-tabs for content sections, and ag-avatar in the header, following the proven AgnosticUI Discovery Dashboard Playbook layout. That is the win.

Agentic Intent is an additive layer on top of the existing playbook system. It does not rewrite your prompt files or change how components work. Each playbook ships with a machine-readable sdui.json file that the CLI reads and injects into your AI context.

How It Works

Every AgnosticUI playbook ships with an sdui.json file that describes:

• Trigger phrases: natural-language expressions that signal the user wants this pattern
• Layout recipe: a plain-English description of the page skeleton
• Component recipe: which AgnosticUI components to use and the role each plays
• Notes: optional implementation hints with small JSON snippets

When you run npx agnosticui-cli context, the CLI scans src/playbooks/ for installed playbooks, reads their sdui.json files, and injects an AgnosticUI Agentic Intent section into your AI context file alongside the component prop references.

npx agnosticui-cli context

If playbooks are present, the CLI announces them:

Detected 2 installed playbooks (dashboard, login) - including intent recipes.

No extra flags are needed. The intent section is injected automatically.

Bundled Playbooks

All eight bundled playbooks ship with sdui.json agentic intent files. Install any of them with:

npx agnosticui-cli playbook <slug> --framework <react|vue|lit>

Then re-run npx agnosticui-cli context to pick up the intent recipes.

Playbook	Slug	Trigger Phrases
Login Form	login	login, sign in, authentication, auth form, user login, sign up
Onboarding Wizard	onboarding	onboarding, wizard, multi-step, getting started, user setup, setup flow, step wizard
Discovery Dashboard	dashboard	dashboard, analytics, metrics, admin panel, overview, data view, home screen
Support Center	support	support, help center, knowledge base, FAQ, tickets, live chat, customer support, help desk
Data Grid	grid	data grid, table, sortable list, data table, inventory, admin table, spreadsheet view
Blog / Article Reader	blog	blog, articles, content listing, post detail, editorial, article reader, reading experience
Landing / Marketing Page	landing	landing page, marketing, hero section, homepage, marketing page, product page, promotional page
Contact Form (FACE)	form-association	form, native form, form submission, form validation, contact form, FACE, form associated custom elements

Authoring a Custom sdui.json

If you have a project-specific playbook, add an sdui.json at the root of its directory (e.g. src/playbooks/my-playbook/sdui.json) to teach your AI assistant how to use it:

{
  "$schema": "https://raw.githubusercontent.com/AgnosticUI/agnosticui/master/v2/playbooks/playbook-sdui.schema.json",
  "version": 1,
  "slug": "my-playbook",
  "displayName": "My Playbook",
  "description": "One-sentence description of what this playbook demonstrates.",
  "intent": {
    "triggers": ["my feature", "my page", "relevant phrase"],
    "summary": "When the user asks for my feature, refer to the My Playbook layout."
  },
  "recipe": {
    "layout": "Plain-English description of the page layout skeleton.",
    "components": [
      { "component": "AgCard", "role": "Primary content container" },
      { "component": "AgButton", "role": "Primary call-to-action" }
    ],
    "notes": "Optional: 1-2 small JSON snippets to give the LLM syntax flavor."
  }
}

The CLI reads any src/playbooks/*/sdui.json it finds. After creating the file, run npx agnosticui-cli context to pick it up.

Required Fields

Field	Type	Description
version	integer	Schema version (currently 1). Increment when the shape changes.
slug	string	Matches the playbook directory name
displayName	string	Human-readable name shown in the injected context
description	string	One-sentence description
intent.triggers	string[]	Natural-language phrases that signal this playbook
intent.summary	string	One-sentence directive for the LLM
recipe.layout	string	Plain-English layout description
recipe.components	array	List of { component, role } objects

Optional Fields

Field	Type	Description
recipe.notes	string	Implementation hints and small JSON snippets for syntax flavor

The version Field and Stale Files

The version field tracks schema compatibility. The current expected version is 1.

If you later update the CLI to a version that expects a higher version number, the CLI warns you:

sdui.json for 'dashboard' uses schema version 1, expected 2 - recipe may be stale.
Re-install: npx agnosticui-cli playbook dashboard --force

Re-installing with --force replaces sdui.json with the latest version while leaving your application code untouched.