Get Shit Done

Stale
GitHub JavaScript MIT

Description

A light-weight and powerful meta-prompting, context engineering, and spec-driven development system for AI coding agents like Claude Code.

Key Features

  • Context engineering framework - Fresh-context subagents solve context rot (quality degradation as context window fills)
  • Five-phase development loop - Disciplined Discuss → Plan → Execute → Verify → Ship iteration cycle
  • Multi-runtime compatible - Supports Claude Code, OpenCode, Gemini CLI, Codex, Cursor, Windsurf and other AI coding tools
  • Structured state persistence - Cross-session memory via structured artifacts like STATE.md and CONTEXT.md
  • Verification-driven delivery - Verify step auto-checks build results and generates fix plans before declaring done

Use Cases

💡 Standardize AI coding agent workflows in large projects to prevent quality degradation from context bloat
💡 Standardize AI-assisted development in team settings with reproducible and auditable development records
💡 Onboard existing codebases to AI-assisted development with reduced integration risk through disciplined processes
💡 Plan and execute complex multi-stage projects with clean context maintained per phase

Strengths & Limitations

Strengths

  • High community interest (64.6k stars)
  • Permissive open-source license (MIT)
  • Responsive to issues, low backlog

Quick Start

# Install GSD Core (interactive runtime selection)
npx @opengsd/gsd-core@latest

# Start a new project
/gsd-new-project

# Enter the five-phase loop: Discuss → Plan → Execute → Verify → Ship

Related Projects

Everything Claude Code

246.4k · JavaScript
Active A+

The agent harness performance optimization system with skills, instincts, memory, security and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

codingagentjavascript +2
  • · Cross-harness compatibility - Works with Codex, Claude Code, Cursor, OpenCode, Gemini, Zed, GitHub Copilot, and more
  • · Skills and instincts system - 63 agents, 251 skills, and 79 legacy command shims as reusable behavior modules
  • · Memory persistence - Hooks automatically save and load context across sessions for continuous memory

Claude Code Game Studios

24.8k · Shell
Stale B

Turn Claude Code into a full game dev studio with 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.

codingagentmulti-agent +1
  • · 49 specialized AI agents — Design, programming, art, audio, narrative, QA and production roles in full studio hierarchy
  • · 73 workflow skills — Slash commands covering every phase from brainstorm to launch
  • · Three-tier studio hierarchy — Directors (Opus), Department Leads (Sonnet), Specialists (Sonnet/Haiku)

AI Website Cloner

33.7k · JavaScript
Active A

Clone any website with one command using AI coding agents. Built with Next.js, React, and shadcn-ui with web scraping and automated code generation.

typescriptagentcoding +3
  • · Clone any website with one command — auto-screenshots, extracts design tokens, generates component specs, and builds a Next.js codebase
  • · Multi-phase pipeline: Recon → Foundation → Component Specs → Parallel Build → Assembly & QA for structured engineering workflow
  • · Supports 13 AI coding agents including Claude Code, Codex CLI, Cursor, Copilot, and more

Scientific Agent Skills

42.1k · Python
Active A+

A set of ready-to-use Agent Skills for research, science, engineering, analysis, finance and writing across multiple coding agents.

agentcodingtools +2
  • · 142 ready-to-use scientific skills — Covers 17+ fields: bioinformatics, cheminformatics, proteomics, clinical research, medical imaging, and more
  • · 100+ scientific database integration — Unified access to 78+ public databases including PubChem, ChEMBL, UniProt, COSMIC, ClinicalTrials.gov
  • · 70+ optimized Python package skills — Dedicated skills for RDKit, Scanpy, PyTorch Lightning, scikit-learn with curated documentation and examples