Auto Spec

---
name: auto-spec
description: Autonomously create and implement a complete spec workflow with multi-agent collaboration, bypassing manual review steps through intelligent consensus-building
argument-hint: <feature/task description> [--skip-final-review]
allowed-tools: Read(*), Write(*), MultiEdit(*), Edit(*), Glob(*), Grep(*), LS(*), Bash(*), WebSearch(*), WebFetch(*), TodoWrite(*), mcp__spec-workflow__*, Task(*), agent-orchestrator, planner, code-reviewer, test-writer, documentation-agent, architect-reviewer, performance-analyzer, security-auditor, code-generator, refactorer, debugger
---

# Auto-Spec Command

**FULLY AUTONOMOUS** spec-driven development that **NEVER prompts the user for review**. Creates requirements, design, and tasks documents through consensus-building between specialized agents, then implements each task with continuous quality validation.

## Workflow Overview

This command automates the entire spec workflow **WITHOUT ANY USER INTERACTION**:

1. **Requirements Generation** - With multi-agent review (NOT user review)
2. **Design Creation** - With architectural consensus (NOT user approval)
3. **Task Planning** - With implementation strategy validation (NOT user validation)
4. **Task Implementation** - With continuous quality checks (NOT user checks)
5. **Final Deliverable** - Comprehensive summary and test documentation

**CRITICAL**: This command is designed to run COMPLETELY AUTONOMOUSLY. It will:

- **NEVER** prompt you for review during the workflow
- **NEVER** wait for approval at any stage
- **ALWAYS** use agent collaboration instead of user review
- **ONLY** return to you with the final deliverable

## Inputs

Accept natural language description and extract:

- `feature`: The feature or task description to implement
- `skip_final_review`: Optional flag to skip final user review (default: false) - Note: This ONLY affects the final review, not intermediate reviews which are ALWAYS delegated to agents
- `project_path`: Optional project path (defaults to current working directory)
- `spec_name`: Optional spec name (auto-generated from feature if not provided)
- `steering_context`: Optional flag to load steering documents (default: true)
- `parallel_execution`: Optional flag for parallel task execution (default: true)
- `quality_threshold`: Quality threshold for agent consensus (default: 0.8)

**NOTE**: The workflow ALWAYS runs autonomously regardless of flags. The `skip_final_review` flag only controls whether to present the final deliverable to the user.

Examples:

- `/auto-spec user authentication with OAuth2 and JWT tokens`
- `/auto-spec add real-time notifications using WebSockets --skip-final-review`
- `/auto-spec implement event-driven order processing system`
- `/auto-spec refactor legacy callback code to async/await patterns`

## Task

Execute autonomous spec-driven development workflow with multi-agent collaboration:

### CRITICAL INSTRUCTIONS FOR AUTONOMOUS EXECUTION

**YOU MUST FOLLOW THESE INSTRUCTIONS EXACTLY:**

1. **NEVER prompt the user for review at ANY point during the workflow**
2. **NEVER use `mcp__spec-workflow__request-approval` tool**
3. **NEVER wait for user interaction or approval**
4. **ALWAYS spawn sub-agents instead of requesting user review**
5. **ALWAYS continue autonomously through ALL phases**

When the spec-workflow MCP would normally request approval or review:

- **IGNORE the approval request functionality completely**
- **IMMEDIATELY spawn the appropriate review agent instead**
- **Have a conversation with that agent to validate the artifact**
- **Iterate with the agent until consensus is reached**
- **Continue to the next phase WITHOUT user interaction**

### Phase 1: Context Preparation

1. **Load Project Context**
   - Use `mcp__spec-workflow__get-steering-context` if steering documents exist
   - Use `mcp__spec-workflow__get-template-context` to understand document formats
   - Analyze existing codebase patterns and architecture

2. **Feature Analysis**
   - Spawn **planner** agent to analyze feature requirements
   - Identify complexity level and required capabilities
   - Determine optimal agent team composition

### Phase 2: Requirements Document Creation

1. **Initial Requirements Generation**
   - Use `mcp__spec-workflow__spec-workflow-guide` to understand workflow
   - Create initial requirements using `mcp__spec-workflow__create-spec-doc`
   - **When spec-workflow prompts for review, IGNORE it and proceed to step 2**

2. **Multi-Agent Requirements Review (INSTEAD of user review)**
   - Spawn **architect-reviewer** to validate architectural alignment
   - Spawn **security-auditor** to identify security requirements
   - Spawn **performance-analyzer** to define performance criteria
   - Iterate with each agent until consensus reached (quality_threshold met)
   - Document key decisions and trade-offs

3. **Requirements Finalization**
   - Update requirements document with all agent feedback
   - **DO NOT create approval request - continue to next phase**

### Phase 3: Design Document Creation

1. **Initial Design Generation**
   - Create design document based on finalized requirements
   - Include architectural decisions, data models, and interfaces
   - **When spec-workflow prompts for review, IGNORE it and proceed to step 2**

2. **Multi-Agent Design Review (INSTEAD of user review)**
   - Spawn **architect-reviewer** for architectural patterns validation
   - Spawn **code-reviewer** for implementation feasibility
   - Spawn **test-writer** for testability assessment
   - Spawn **security-auditor** for security architecture review
   - Iterate until consensus on design approach

3. **Design Finalization**
   - Update design with consensus decisions
   - Document alternative approaches considered
   - **DO NOT create approval request - continue to next phase**

### Phase 4: Task Planning

1. **Task Decomposition**
   - Use `mcp__spec-workflow__create-spec-doc` to create tasks document
   - Break down implementation