prompt-search

name	prompt-search
description	Find, retrieve, and improve AI prompt templates using GitHub MCP search patterns and web fallbacks — no registry MCP server required. Use when the user asks about prompt templates, wants to search prompts, or improve a prompt.

Prompt Search

Locate prompt templates and prompt libraries using GitHub search and the web.

Workflow

1. GitHub MCP (primary — there is no dedicated CLI for prompts):
   • github_text_search with path:prompts extension:md <keyword>, scoped to repos with prompt dirs (github/awesome-copilot, f/prompts.chat, or relevant orgs).
   • Browse curated prompt collections directly and pull raw files.
2. Web/browser fallback: fetch prompts.chat or awesome lists; present title, description, author, category, and link.

How to Retrieve / Improve

• Fetch the raw file (GitHub MCP or web) and check for variables (${variable} / ${variable:default}) — ask the user for values when present.
• Improve a prompt by rewriting while preserving intent; explain what was enhanced.
• Install by saving as .prompt.md in the workspace prompts dir (team-shared) or the user prompts folder (personal); keep it parameterized and generic.

Guidelines

• Always search before writing a prompt from scratch.
• Show readable results with links and categories.
• When improving, explain what was enhanced.
• Suggest relevant categories/tags when saving.