mad-full

What problem does it solve?

Manually coordinating the full feature development lifecycle from initial idea to merged pull request is error-prone, inconsistent, and time-consuming, requiring teams to manually manage specifications, plans, task tracking, implementation, testing, and PR reviews across multiple disconnected steps. This Skill eliminates that overhead by automating the entire MAD pipeline with built-in quality guardrails and error recovery.

Core Features & Use Cases

• Full Pipeline Orchestration: Automates the complete end-to-end feature development workflow (specify → plan → tasks → implement → validate → commit → PR) with mandatory quality review gates at each phase to catch issues early.
• Adaptive Workflow Depth: Automatically triages feature complexity (trivial, standard, complex) to skip unnecessary steps for small changes like typo fixes, while running deep research for architectural changes.
• Parallel Agent Coordination: Leverages Agent Teams to run parallel validation lenses, implementation workers, and research scouts to speed up workflow execution without manual coordination.
• Multiple Execution Modes: Supports interactive mode with user checkpoints for learning, autonomous mode for unattended overnight runs, and headless JSON output for CI/CD integration.
• Error Recovery: Includes built-in recovery logic for test failures, build errors, and gate failures, with automatic retry logic and progress checkpoints to resume interrupted runs.
• Use Case: A developer can provide a natural language request like "add OAuth2 user authentication" and automatically generate the full specification, implementation plan, task list, write and test the code, and open a PR for review without manually coordinating each step.

Quick Start

Use the mad-full skill by providing your natural language feature description to automatically run the complete end-to-end development pipeline from specification through merged pull request.