gsd

What This Skill Does

GSD (Get Shit Done) is a comprehensive project management and execution engine for OpenClaw. It serves as a full-cycle workflow architect that migrates complex projects from abstract ideas into tangible, verified deliverables. The skill utilizes a multi-agent framework to handle deep context gathering, parallelized domain research, and step-by-step roadmap generation. By breaking development into manageable phases and execution waves, it ensures that every line of code is intentional, researched, and verified against defined goals. GSD is designed to minimize scope creep and context loss, providing a robust persistent state that allows developers to pause, resume, and audit work at any stage of the project lifecycle.

Installation

To integrate GSD into your environment, run the following command within your OpenClaw terminal: clawhub install openclaw/skills/skills/oleg-schmidt/gsd

Use Cases

• Brownfield Projects: Use map-codebase to analyze legacy code and systematically plan migrations or feature additions.
• Greenfield Development: Kickstart new projects with new-project to automatically generate requirements, architectures, and phase-based roadmaps.
• Complex Feature Delivery: Implement large-scale updates by leveraging phase-based planning and automated verification loops.
• Debugging & Maintenance: Use the debug command for systematic, persistent troubleshooting that tracks state across sessions.

Example Prompts

• "/gsd new-project - I want to build a decentralized task manager with a React frontend and a Rust backend."
• "/gsd plan-phase 2 --research - Research the best authentication libraries for this stack and draft the execution plan for user login."
• "/gsd progress - Where are we in the current milestone, and what is the next critical task I should focus on?"

Tips & Limitations

• Trust the Workflow: GSD is designed for high-ceremony, high-quality output. Don't skip the discuss-phase if you aren't sure about your requirements; it provides essential clarity.
• Use Flags Wisely: Use --skip-research only for tasks where you already possess domain expertise to save time, but default to full research for unfamiliar technologies.
• Verification is Key: Always use verify-work after execute-phase to ensure the agent’s logic aligns with your vision.
• Persistence: The skill tracks state, so if you get interrupted, use pause-work to capture your current progress properly before closing the session.