SkillsApr 7, 2026·3 min read

Get Shit Done (GSD) — Meta-Prompting Dev System for Claude Code

A spec-driven development system with 48.6k GitHub stars. Adds phase-based planning, multi-agent execution, verification gates, and state persistence to Claude Code, Cursor, Gemini CLI and 9 more runtimes. Install with one npx command.

HE
henuwangkai · Community
Quick Use

Use it first, then decide how deep to go

This block should tell both the user and the agent what to copy, install, and apply first.

npx get-shit-done-cc --claude --global

Restart Claude Code, then run /gsd:new-project to start.

For other runtimes:

npx get-shit-done-cc --cursor --global   # Cursor
npx get-shit-done-cc --gemini --global   # Gemini CLI
npx get-shit-done-cc --codex --global    # Codex
npx get-shit-done-cc --windsurf --global # Windsurf
Intro

Get Shit Done (GSD) is a meta-prompting, context engineering and spec-driven development system with 48,600+ GitHub stars. It solves the "context rot" problem in AI coding by organizing work into structured phases — discussion → planning → execution → verification — each with fresh context windows. Supports 12 runtimes including Claude Code, Cursor, Gemini CLI, Codex, Windsurf, and Trae.

Best for: developers using AI coding assistants on non-trivial projects. Works with: Claude Code, Cursor, Gemini CLI, Codex, Windsurf, Copilot, Augment, Trae, Cline, Kilo, Antigravity, OpenCode. Setup time: under 1 minute.

Core Features

Phase-Based Workflow

GSD breaks complex projects into numbered phases, each going through a structured cycle:

  1. Discuss (/gsd:discuss-phase) — Adaptive questioning to gather context
  2. Plan (/gsd:plan-phase) — Detailed plan with verification criteria
  3. Execute (/gsd:execute-phase) — Multi-agent parallel execution with atomic commits
  4. Verify (/gsd:verify-work) — Automated verification against success criteria

Multi-Agent Orchestration

  • Spawns parallel sub-agents with fresh context windows
  • Wave-based execution respects file dependencies
  • Each agent gets exactly the context it needs — no bloat

Quality Gates

  • Plan-checker validates plans before execution
  • Schema drift detection catches ORM/migration mismatches
  • Security enforcement via /gsd:secure-phase
  • Scope reduction detection prevents silent requirement drops

State Persistence

  • PROJECT.md — Project decisions and architecture
  • ROADMAP.md — Phase roadmap with progress tracking
  • STATE.md — Current execution state
  • Survives context resets and session restarts

68+ Built-in Commands

  • /gsd:new-project — Initialize project with deep context gathering
  • /gsd:autonomous — Run all phases autonomously
  • /gsd:debug — Systematic debugging with persistent state
  • /gsd:quick — Quick tasks with GSD guarantees
  • /gsd:fast — Trivial tasks, no overhead
  • /gsd:stats — Project statistics and timeline
  • /gsd:ship — Create PR and prepare for merge

Codebase Intelligence (v1.33+)

Queryable .planning/intel/ store with structured JSON files for files, exports, symbols, patterns, and dependencies. Incremental updates keep it current.

FAQ

Q: What is GSD (Get Shit Done)? A: GSD is an open-source meta-prompting system that adds structured project management, multi-agent execution, and quality verification to AI coding tools like Claude Code and Cursor. 48.6k GitHub stars, MIT licensed.

Q: Is GSD free? A: Yes, fully open-source under the MIT license. Install via npm.

Q: How do I install GSD? A: Run npx get-shit-done-cc --claude --global and restart Claude Code. Takes under 1 minute.

Q: What runtimes does GSD support? A: Claude Code, Cursor, Gemini CLI, Codex, Windsurf, Copilot, Augment, Trae, Cline, Kilo, Antigravity, and OpenCode — 12 runtimes total.

🙏

Source & Thanks

Created by TÂCHES. Licensed under MIT.

get-shit-done — ⭐ 48,600+

A game-changing development workflow system for the AI coding era. Join the community on Discord.

Discussion

Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.

Related Assets

CC Status Board

Smart status bar for Claude Code — context meter, AI asset discovery with bilingual matching, and session info at a glance. v1.1: Uses claude CLI to auto-generate Chinese keywords at index time. Zero config, zero extra API keys.

henuwangkai
◈Home🔍Search👤Me