# Smart Prompts MCP Server - Project Plan
## Overview
An enhanced MCP server that fetches prompts from GitHub repositories and provides intelligent prompt discovery, composition, and management features.
## Architecture
### Core Components
1. **GitHub Integration**
- Fetch prompts directly from GitHub repositories (public/private)
- No local storage required
- Watch for updates and cache with periodic refresh
2. **MCP Protocol Implementation**
- **Resources** (Application-controlled context)
- **Prompts** (User-controlled templates)
- **Tools** (Model-controlled actions)
### MCP Resources
- `prompts://list` - List all available prompts
- `prompts://category/{name}` - Get prompts by category
- `prompts://search?q={query}` - Search prompts
- `prompts://stats` - Usage statistics
- `prompts://recent` - Recently used prompts
### MCP Prompts
Dynamic prompt templates with:
- Arguments support (required/optional)
- Context-aware suggestions
- Multi-step workflows
- Prompt composition capabilities
### MCP Tools
- `fetch_prompt` - Get specific prompt from GitHub
- `search_prompts` - Semantic search across prompts
- `compose_prompt` - Combine multiple prompts
- `suggest_prompts` - AI-powered suggestions
## Prompt File Structure
```yaml
---
name: code_review_checklist
category: coding
tags: [review, quality, best-practices]
description: Comprehensive code review template
arguments:
- name: language
required: true
description: Programming language
- name: focus_areas
required: false
description: Specific areas to focus on
---
# Code Review Checklist for {{language}}
## General Guidelines
- [ ] Code follows project style guidelines
- [ ] Functions and variables have descriptive names
- [ ] Complex logic is well-commented
{{#if focus_areas}}
## Focus Areas: {{focus_areas}}
{{/if}}
[Additional prompt content...]
```
## Enhanced Features
1. **Smart Discovery**
- Semantic search using embeddings
- Context-aware suggestions based on current task
- Related prompt recommendations
2. **Dynamic Composition**
- Combine multiple prompts
- Fill templates with context
- Chain prompts for workflows
3. **Usage Intelligence**
- Track most used prompts
- Suggest improvements
- Learn from usage patterns
4. **GitHub Integration**
- Fetch directly from repo (no local storage)
- Watch for updates
- Support private repos with auth
## Implementation Example
```typescript
// Server structure
{
resources: {
// Expose prompt data
"prompts://list": getAllPrompts(),
"prompts://category/coding": getPromptsByCategory("coding"),
"prompts://prompt/{id}": getPromptById(id)
},
prompts: {
// Interactive prompt templates
"enhance_code": {
description: "Enhance code with best practices",
arguments: [
{ name: "language", required: true },
{ name: "framework", required: false }
]
},
"generate_tests": {
description: "Generate test cases",
arguments: [
{ name: "test_framework", required: true }
]
}
},
tools: {
// Actions the AI can take
"search_prompts": searchWithEmbeddings,
"compose_prompt": combinePrompts,
"analyze_context": suggestRelevantPrompts
}
}
```
## Benefits
- No local file management required
- Rich MCP protocol features (resources, prompts, tools)
- Intelligent prompt discovery and composition
- Better integration with AI assistants
- Community contribution potential
- Version control through GitHub
- Easy sharing and collaboration
