spec_driven_develop
Spec-driven workflow plugin for AI coding agents with GitHub issue tracking.
Architecture-first planning, task decomposition, and adaptive control for Claude Code and Cursor.
Spec-Driven Develop is a pure Markdown workflow plugin that brings architecture-first planning and task decomposition to AI coding agents like Claude Code, Cursor, and Codex. It suits you if you need to manage complex code migrations or rewrites without relying on heavy third-party runtimes.
What it is
It guides AI coding agents through a seven-phase pipeline ranging from quick intent capture and deep project analysis to task decomposition and progress tracking. The tool automatically detects your environment to handle GitHub issues, milestones, and batched pull requests while applying engineering cybernetics to correct plan drift.
Who it's for
It suits software engineers and architects orchestrating large-scale refactors, migrations, or multi-file code changes using Markdown-compatible AI agents. It does not suit developers looking for simple point-and-click extensions or those who do not use AI coding assistants.
Key features
Seven-Phase Pipeline
Automates tasks from intent capture and deep analysis to task decomposition and archiving.
GitHub Integration
Automatically creates issues, milestones, and delivery batch pull requests based on environment detection.
Adaptive Control
Implements a closed-loop feedback system to monitor execution telemetry and correct plan drift.
Tiered Execution
Supports orchestrator-direct execution, single-coder tasks, and parallel lanes with dedicated review agents.
Up and running in 3 steps
Our honest verdict
What's great
- Pure Markdown workflows with zero SDK or third-party runtime dependencies
- Auto-detects GitHub environments to manage issues, labels, and milestones
- Includes a closed-loop feedback mechanism to correct drift during complex tasks
- Provides structured deep-discussion and findings-first code review capabilities
Where it falls short
- Requires familiarity with command-line AI coding agents like Claude Code or Cursor
- Running automated parallel agent lanes and managing GitHub workflows demands active oversight
- Relies entirely on the capabilities of the underlying Markdown-compatible AI agent
Pricing & license
Free under its license, with no paid hosted option since it runs locally inside your AI coding agent environment. The only cost is your time and AI token usage.
Open remote AI-training roles
873 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
What AI agents are supported?
It works with Claude Code, Codex, OpenCode, Cursor, Windsurf, Cline, Aider, Continue, Roo Code, Augment, and any agent that reads custom Markdown skills.
Does it require external dependencies?
No, it consists of pure Markdown workflows and small helper scripts with zero runtime dependencies.
How does task tracking work?
If a GitHub repository is detected, it automatically creates issues, milestones, and labels, falling back to local Markdown tracking otherwise.
