GSD for Antigravity Get Shit Done addresses the unreliability of AI-generated code by providing a structured, context-engineered development methodology. It targets solo developers and small teams seeking consistent, high-quality code output without the overhead of traditional agile processes. The benefit is a reliable system for building software with AI assistance.

How It Works

GSD employs "context engineering" to feed AI models precise project information via structured files like SPEC.md and ARCHITECTURE.md. Development proceeds through distinct phases: Initialize (questioning to finalize SPEC.md), Discuss (clarify scope), Plan (generate PLAN.md with XML-formatted tasks), Execute (wave-based, parallelized task execution), and Verify (empirical evidence collection). XML task formatting ensures unambiguous instructions and built-in verification steps, while wave-based execution and atomic Git commits maintain clean context and granular control, preventing AI hallucinations and ensuring traceability.

Quick Start & Requirements

To begin, clone the repository (git clone https://github.com/toonight/get-shit-done-for-antigravity.git) and copy the template files (.agent, .gemini, .gsd) into your project directory. Subsequently, initiate a new project using the /new-project command within the agent. Prerequisites include PowerShell (Windows) or Bash (Linux/Mac) and Git. No specific hardware, CUDA, or Python versions are mandated by the README.

Highlighted Details

• Context Engineering: Manages critical project documents (SPEC.md, ARCHITECTURE.md, STATE.md, etc.) to provide AI with a consistent "world model."
• XML Prompt Formatting: Defines tasks with explicit actions, verification steps, and expected outcomes, optimizing AI execution.
• Wave-Based Execution: Organizes tasks into dependency-based waves, enabling parallelization and ensuring fresh context for each execution cycle.
• Atomic Git Commits: Each task results in an individual commit, facilitating precise debugging, revertibility, and a clear project history.
• Empirical Verification: Mandates proof of completion (e.g., command output, screenshots) for every task, ensuring demonstrable correctness.

Maintenance & Community

The project is adapted from glittercowboy/get-shit-done. No specific details regarding active maintainers, community channels (like Discord or Slack), or sponsorship are provided in the README.

Licensing & Compatibility

The README does not specify a software license. This absence creates ambiguity regarding usage rights, modification, and distribution, particularly for commercial applications.

Limitations & Caveats

The methodology imposes size limits on context files to maintain AI quality. The project's origin as an adaptation for "Google Antigravity" may imply specific, unstated constraints or a focus on internal Google workflows. Crucially, the lack of a specified license poses a significant adoption blocker.