Expert-level feature planning workflow with integrated research (Context7 + Perplexity)
Transform vague feature requests into comprehensive, research-backed implementation plans with automated Context7 documentation lookup and Perplexity industry research.
UltraPlan is a systematic planning workflow that:
- Researches official documentation (Context7) and industry best practices (Perplexity)
- Analyzes your project structure and existing patterns
- Creates comprehensive planning documents with user stories, acceptance criteria, and task breakdowns
- Clarifies requirements through interactive questions
- Implements features following the approved plan with incremental testing
- 🔍 Automated Research - Context7 for official docs, Perplexity for FAANG patterns and community practices
- 📋 Comprehensive Planning - Requirements, user stories, implementation approaches, risk assessment
- 🤔 Interactive Clarification - Smart questions based on feature complexity
- ✅ Task Breakdown - Detailed, actionable tasks with time estimates
- 📝 Living Documentation - Plans evolve during implementation with notes and learnings
- 🧪 Testing Strategy - Includes test planning and Playwright E2E guidance
- Claude Code CLI - Install from claude.com/claude-code
- MCP Servers (required for full functionality):
- Context7 MCP - Official documentation research
- Perplexity MCP - Industry research
# Install from GitHub
claude skill install github:anombyte93/ultraplan-skill
# Or install from local directory
git clone https://github.com/anombyte93/ultraplan-skill.git
claude skill install ./ultraplan-skillAdd to your .mcp.json or ~/.claude/mcp.json:
{
"mcpServers": {
"context7": {
"command": "npx",
"args": ["-y", "@context7/mcp-server"]
},
"perplexity-free": {
"command": "npx",
"args": ["-y", "@perplexity-mcp/server"],
"env": {
"PERPLEXITY_API_KEY": "your_api_key_here"
}
}
}
}Get API Keys:
- Context7: No API key needed (uses npx)
- Perplexity: Get from perplexity.ai/settings/api
# From Claude Code
/ultraplan Build a user authentication system with OAuth/ultraplan Build a real-time collaborative editing feature with conflict resolution using operational transforms. Should integrate with our existing WebSocket infrastructure and support 10+ concurrent users per document.- Planning Mode Activates - Claude enters planning mode
- Research Phase - Queries Context7 and Perplexity for best practices
- Project Analysis - Examines your codebase structure
- Document Creation - Creates comprehensive plan in
.taskmaster/docs/plans/pending/ - Clarifying Questions - Asks 3-4 questions about scope, approach, timeline, testing
- Approval - Asks if you're ready to implement
- Implementation - Creates TodoWrite checklist and begins coding
Plans are saved in .taskmaster/docs/plans/:
.taskmaster/docs/plans/
├── pending/
│ └── user-auth-plan-2025-10-29.md # Active planning
└── complete/
└── user-auth-plan-2025-10-29.md # Implemented features
Each plan includes:
- Executive Summary - Quick overview
- Requirements - Functional and non-functional requirements
- User Stories - With acceptance criteria
- Research Findings - Industry standards, community practices, official docs
- Implementation Approaches - Multiple options with pros/cons
- Tools & Resources - Libraries, frameworks, project utilities
- Success Criteria - Definition of done and quality gates
- Risk Assessment - Potential issues and mitigation strategies
- Task Breakdown - Detailed, actionable tasks with estimates
- References - All research sources
By default, plans are saved to .taskmaster/docs/plans/. To customize:
Edit skill.json in your local skill installation:
{
"configuration": {
"default_plan_directory": "docs/planning"
}
}The skill asks 3-4 questions by default. You can modify main.md to add custom questions for your team's workflow.
/ultraplan Add infinite scroll to the product listing page with lazy loading and skeleton statesResult:
- Researches React infinite scroll patterns
- Checks Intersection Observer API docs
- Finds FAANG performance best practices
- Creates plan with performance metrics
- Asks about pagination strategy and UX preferences
/ultraplan Implement rate limiting middleware with Redis for our Express APIResult:
- Researches Express middleware patterns
- Checks Redis rate limiting strategies
- Finds industry standard rate limit algorithms (token bucket, sliding window)
- Creates plan with scalability considerations
- Asks about rate limit thresholds and error responses
/ultraplan Build a CLI tool for database migrations with rollback supportResult:
- Researches migration tool patterns (Alembic, Flyway, etc.)
- Checks Node.js CLI best practices
- Finds database versioning strategies
- Creates plan with migration file structure
- Asks about database targets and rollback scenarios
- Specific: "Add JWT authentication with refresh tokens"
- Contextual: "...integrate with our existing user service"
- Scoped: "...support email/password and Google OAuth"
- Constrained: "...must work with PostgreSQL user table"
"Add authentication"→ Too broad"Make it faster"→ No measurable goal"Fix the bug"→ Not a feature plan
- Mention existing patterns - "...following our repository pattern"
- State priorities - "...prioritize developer experience over performance"
- Include constraints - "...must not require database migration"
- Specify scale - "...support 1000 concurrent users"
Install Context7 MCP:
# Add to .mcp.json
{
"mcpServers": {
"context7": {
"command": "npx",
"args": ["-y", "@context7/mcp-server"]
}
}
}Get a Perplexity API key and add to .mcp.json:
{
"mcpServers": {
"perplexity-free": {
"command": "npx",
"args": ["-y", "@perplexity-mcp/server"],
"env": {
"PERPLEXITY_API_KEY": "pplx-xxxxx"
}
}
}
}UltraPlan will create .taskmaster/docs/plans/ automatically. If you prefer a different location, update skill.json.
Check your MCP server logs:
# In Claude Code, enable MCP debug mode
claude --mcp-debugContext7 Queries:
- Identifies libraries/frameworks in feature request
- Resolves library IDs automatically
- Fetches official documentation with focused topics
- Extracts code examples and best practices
Perplexity Queries (3 queries):
- Industry Standards - FAANG companies, production patterns
- Community Practices - GitHub trends, popular libraries
- Current Research - 2024-2025 papers, conference talks
The skill uses research findings to:
- Recommend proven patterns over experimental approaches
- Suggest libraries with strong community adoption
- Identify performance bottlenecks before coding
- Propose test strategies based on feature complexity
| Traditional | UltraPlan |
|---|---|
| Manual research | Automated Context7 + Perplexity |
| Static documents | Living docs with implementation notes |
| Vague requirements | User stories with acceptance criteria |
| Unknown risks | Research-backed risk assessment |
| No task breakdown | Detailed, actionable tasks |
- More comprehensive than simple task breakdown
- Research-backed unlike template-based planning
- Interactive with clarifying questions
- Integrated with TaskMaster workflow
Found a bug or have a suggestion?
- Open an issue: https://github.com/anombyte93/ultraplan-skill/issues
- Submit a PR with improvements
- Share your experience in discussions
MIT License - see LICENSE file
Created by @anombyte93
Powered by:
- Initial release
- Context7 and Perplexity integration
- Interactive clarification questions
- Comprehensive planning documents
- Task breakdown with time estimates
- Living documentation during implementation
Ready to plan like an expert?
claude skill install github:anombyte93/ultraplan-skill
cd your-project
/ultraplan Build an amazing feature