Cursor Agentic Development Pipeline

A comprehensive, AI-powered software development workflow that orchestrates multiple specialized agents through a structured pipeline to deliver complete epics from conception to deployment.

Overview

This system implements a 5-phase agentic pipeline that transforms high-level ideas into production-ready software through coordinated AI agents. Each agent has a specialized role and operates within strict constraints to ensure quality, consistency, and traceability.

Architecture

Core Pipeline Structure

Foundation & Setup (00-03)
├── commands/00-start.md       - Initialize new epic
├── support/01-forbidden.md    - Universal constraints & guardrails
├── support/02-mcp.md          - MCP server activation
├── support/03-pipeline.md     - Pipeline orchestration
└── support/04-rules.md        - Rule loader

Planning Phase (10-17)
├── 11-discuss.md   - Product discovery
├── 12-idea.md      - Concept development
├── 13-packages.md  - Dependency analysis
├── 14-prd.md       - Requirements specification
├── 15-user-stories.md - User journey mapping
├── 16-mermaid.md   - Architecture visualization
└── 17-create.md    - Task breakdown

Role Definition (20-26)
├── 21-devops.md    - Infrastructure planning
├── 22-architect.md - System design
├── 23-backend.md   - Backend implementation
├── 24-frontend.md  - Frontend implementation
├── 25-test.md      - Testing strategy
└── 26-lead.md      - Project leadership

Process Management (30-31)
└── 31-expand.md    - Task expansion & refinement

Development Phase (40-44)
├── 41-open.md      - Development initiation
├── 42-code.md      - Implementation
├── 43-verify.md    - Quality assurance
└── 44-close.md     - Task completion

Finalization (50-54)
├── 51-docs.md      - Documentation
├── 52-qa.md        - Quality audit
├── 53-done.md      - Epic completion
└── 54-pr.md        - Pull request creation

PR Review Phase (60-64)
├── 61-review.md    - PR review analysis
├── 62-address.md   - Review feedback implementation
├── 63-verify.md    - Review changes verification
└── 64-commit.md    - Review changes commit & push

Research Phase (80-82)
├── 81-research.md  - Package research
└── 82-gap.md       - Gap analysis

Operations (99)
└── 99-rebake.md    - Re-bake epic artifacts

📖 The command files shown in the pipeline structure above correspond to the agent roles listed below. Click any command number in the agent roles section to view detailed implementation.

MCP Server Integrations

The system leverages multiple Model Context Protocol (MCP) servers for enhanced capabilities:

🤖 Task Master AI (task-master-ai)

Multi-provider AI orchestration supporting:

• Anthropic Claude
• OpenAI GPT
• Google Gemini
• Mistral
• Perplexity
• Groq
• Azure OpenAI
• Ollama

📚 Context7 (context7)

• Library and documentation integration
• Real-time API references
• Framework documentation

🔍 Perplexity (perplexity)

• Web research and analysis
• Real-time information gathering
• Technical research capabilities

🌐 DeepWiki (deepwiki)

• Repository analysis and insights
• Code pattern recognition
• Best practices research

🛠️ Laravel Boost (laravel-boost)

• Laravel framework expertise
• PHP ecosystem knowledge
• Laravel-specific optimization

🧠 Sequential Thinking (sequential-thinking)

• Structured reasoning processes
• Step-by-step problem solving
• Logic validation and refinement

🗂️ New Knowledge (knowledgegraph)

• Knowledge graph management
• Memory and context retention
• Related information discovery

🐙 GitHub (github)

• GitHub integration
• Repository management
• PR and issue handling

Agent Roles & Responsibilities

🎯 Planning Phase (10-17)

• Product Manager (11-discuss): Problem space definition and user research
• Innovation Lead (12-idea): Creative solution ideation
• Package Manager (13-packages): Technology stack analysis
• Requirements Engineer (14-prd): Technical specification writing
• UX Researcher (15-user-stories): User journey and story mapping
• System Architect (16-mermaid): Visual system design
• Project Manager (17-create): Task decomposition and planning

👥 Role Definition Phase (20-26)

• DevOps Engineer (21-devops): Infrastructure and deployment planning
• System Architect (22-architect): Technical architecture design
• Backend Developer (23-backend): API and data layer design
• Frontend Developer (24-frontend): User interface planning
• QA Engineer (25-test): Testing strategy and validation
• Engineering Lead (26-lead): Technical leadership and oversight

⚙️ Process Phase (30-31)

• Process Engineer (31-expand): Task refinement and optimization

💻 Development Phase (40-44)

• Development Lead (41-open): Implementation coordination
• Software Engineer (42-code): Code implementation and integration
• Quality Engineer (43-verify): Testing and validation
• Release Manager (44-close): Feature completion and handover

📋 Finalization Phase (50-54)

• Technical Writer (51-docs): Documentation creation
• Quality Auditor (52-qa): Comprehensive quality review
• Completion Lead (53-done): Epic completion and knowledge capture
• Integration Lead (54-pr): Pull request and deployment preparation

🔀 PR Review Phase (60-64)

• Code Review Engineer (61-review): PR review analysis and documentation
• Development Engineer (62-address): Review feedback implementation
• Quality Engineer (63-verify): Review changes verification
• Release Engineer (64-commit): Review changes commit and push

🔬 Research Phase (80-82)

• Package Research Partner (81-research): Laravel package research and evaluation
• Gap Analysis Partner (82-gap): Comparative analysis between solutions

🔄 Operations (99)

• Planning Steward (99-rebake): Re-bake epic artifacts and harmonize existing planning

Universal Constraints & Guardrails

All agents operate under strict support/01-forbidden.md constraints:

✅ Allowed Actions

• Truth over assumption: All data must trace to PRD, tasks, or user input
• Local-first operations: All work within project workspace
• Non-destructive changes: No modification of unrelated components
• MCP-orchestrated actions: All operations through approved tools
• Rule-based execution: Commands must load shared rules from support/04-rules.md

❌ Forbidden Actions

• Time estimation or scope invention
• Code generation during planning phases
• External dependency addition without approval
• Security credential exposure
• Unapproved documentation modification
• Direct manipulation of task-master files (must use MCP tools)
• Speculative recommendations or "Next steps" in output

🎯 Quality Gates

• Pint: Code cleanliness and standards
• Larastan: Static analysis (level ≥6 or baseline)
• Mutation Testing: ≥70% coverage
• Performance: p95 ≤500ms, ≤10 queries
• Error Tracking: 0 unhandled exceptions

📋 Support Files

All commands reference these foundational files:

• support/01-forbidden.md - Universal constraints and guardrails
• support/02-mcp.md - MCP server activation and usage
• support/03-pipeline.md - Pipeline orchestration guide
• support/04-rules.md - Rule loader for epic workflows

Getting Started

Prerequisites

1. Cursor with MCP server support
2. API Keys for integrated services (see mcp.json)
3. GitHub repository access
4. Laravel project environment (optional but recommended)

Configuration

1. Setup MCP Servers: Configure mcp.json with your API keys
2. Review Support Files: All commands reference support/01-forbidden.md, support/02-mcp.md, support/03-pipeline.md, and support/04-rules.md
3. Initialize Epic: Run /00-start "Your Epic Title"
4. Follow Pipeline: Execute commands sequentially through each phase

Example Workflow

# Initialize new epic
00-start

# Planning phase
11-discuss
12-idea
13-packages
14-prd
15-user-stories
16-mermaid
17-create

# Role definition
21-devops
22-architect
23-backend
24-frontend
25-test
26-lead

# Process refinement
31-expand

# Development execution
41-open
42-code
43-verify
44-close

# Finalization
51-docs
52-qa
53-done
54-pr

# PR Review (when feedback received)
61-review
62-address
63-verify
64-commit

# Research (as needed)
81-research  # Package research
82-gap       # Gap analysis

# Operations (as needed)
99-rebake

Key Features

🔄 Automated Orchestration

• Sequential command execution
• Context preservation across phases
• Automatic artifact generation
• Git integration and tracking

🎯 Specialized Intelligence

• Role-specific AI agents
• Multi-provider AI integration
• Contextual knowledge retention
• Research and analysis capabilities

🛡️ Quality Assurance

• Built-in constraint enforcement
• Multi-stage validation
• Comprehensive testing integration
• Performance monitoring

📊 Traceability

• Complete audit trails
• PRD-to-implementation mapping
• Task and artifact linking
• Git commit integration

Integration Examples

GitHub Integration

• Automatic branch creation
• PR generation with descriptions
• Issue tracking and linking
• Repository analysis

Documentation Generation

• API documentation
• User guides
• Architecture diagrams
• Deployment guides

Testing Integration

• Unit test generation
• Integration testing
• Performance testing
• Quality gate validation

Advanced Features

🔄 Context Management

• Knowledge graph integration
• Memory retention across sessions
• Related epic discovery
• Pattern recognition

🔍 Research Capabilities

• Real-time web research
• Library and framework analysis
• Best practice identification
• Technical trend awareness

🎨 Visualization

• Architecture diagrams (Mermaid)
• User journey mapping
• System flow documentation
• Dependency visualization

Contributing

This system is designed as a complete, self-contained workflow. Each command file represents a specific agent role and should maintain:

1. Single responsibility per command
2. Clear input/output contracts
3. MCP tool integration
4. Constraint compliance with support/01-forbidden.md
5. Comprehensive documentation

File Structure

.cursor/
├── commands/          # Command implementations organized by phase
│   ├── 00-start.md
│   ├── 10-planning/
│   ├── 20-roles/
│   ├── 30-process/
│   ├── 40-dev/
│   ├── 50-final/
│   ├── 60-git/
│   ├── 80-research/
│   └── 99-rebake.md
├── support/           # Universal constraints and configuration
│   ├── 01-forbidden.md
│   ├── 02-mcp.md
│   ├── 03-pipeline.md
│   └── 04-rules.md
├── rules/             # Epic-specific rules (epics/*.mdc)
│   └── epics/
└── mcp.json           # MCP server configuration

License

This agentic workflow system is designed for Laravel ecosystem development but can be adapted for other frameworks with appropriate MCP server configuration.

Key Workflow Features

🔄 Re-baking Epics

The 99-rebake command allows you to refactor and harmonize existing epic planning artifacts without creating new scope. Useful when requirements evolve or artifacts become stale.

📝 Epic Completion

The 53-done command finalizes completed epics by:

• Synthesizing delivery summaries
• Updating epic task records
• Recording knowledge graph insights
• Marking epics as complete

---

Built for Cursor • MCP-Powered • Quality-First • Traceable Development