Skip to content
starlight-llm-actions

Example: Per-provider prompt

This page combines a page-level prompt and a per-provider prompt for Cursor specifically — the deepest layer of the precedence model.

What’s set

Section titled “What’s set”
this page's frontmatter
---
title: 'Example: Per-provider prompt'
llmActions:
  prompt: 'General context for this page: {md_url}'
  actions:
    openIn:
      providers:
        cursor:
          prompt: 'Use this docs page to scaffold an Acme SDK integration in a fresh project.'
---

What you should see

Section titled “What you should see”
  • Click Open in… → Cursor: the prompt sent is the SDK-scaffolding one.
  • Click Open in… any other provider (ChatGPT, Claude, Gemini, …): the prompt sent is the page-level “General context for this page” one.

When to use this

Section titled “When to use this”
  • Code editors get coding prompts. Cursor / Copilot benefit from a prompt focused on integration code; general-purpose chats benefit from a conversational prompt.
  • Demos vs. reference. A demo page might prompt Cursor to “build the example in a new project” while telling ChatGPT “explain this page line by line.”

How precedence works (most-specific wins)

Section titled “How precedence works (most-specific wins)”
  1. Per-page + per-provider prompt — what Cursor uses on this page.
  2. Per-page top-level prompt — what every other provider uses on this page.
  3. Global per-provider prompt — would be used if no per-page prompt were set for that provider.
  4. Global top-level prompt — final fallback.