Get Shit Done (GSD): The Complete Guide to Meta-Prompting & Context Engineering for Claude Code
Vibecoding has a bad reputation: describe what you want, AI generates code, you get inconsistent garbage that falls apart at scale. Get Shit Done (GSD) fixes that. It's the context engineering layer that makes Claude Code reliable — meta-prompting, multi-agent orchestration, XML-structured plans, atomic git commits, and spec-driven development, all wrapped in simple slash commands. With 26,000+ GitHub stars, GSD is the #1 system for building real software with AI coding agents.
What Is GSD?
GSD is a lightweight, powerful meta-prompting, context engineering and spec-driven development system originally built for Claude Code, now also supporting OpenCode, Gemini CLI, and Codex. Created by TÂCHES, a solo developer who doesn't write code — Claude Code does.
- Language: JavaScript
- License: MIT
- Stars: 26,000+ ⭐
- Forks: 2,200+
- Releases: 32
- Install:
npx get-shit-done-cc@latest
The Philosophy: No Enterprise Theater
"Other spec-driven development tools exist; BMAD, Speckit... But they all seem to make things way more complicated than they need to be. I'm not a 50-person software company. I don't want to play enterprise theater." — TÂCHES
GSD's principle: the complexity is in the system, not in your workflow. Behind the scenes: context engineering, XML prompt formatting, subagent orchestration, state management. What you see: a few commands that just work.
The 6-Phase Workflow
1. Initialize Project
/gsd:new-project [--auto]
GSD extracts everything it needs to know about your idea.
2. Discuss Phase
/gsd:discuss-phase [N] [--auto]
Deep discussion to clarify requirements, explore edge cases, and research.
3. Plan Phase
/gsd:plan-phase [N] [--auto]
Generates XML-structured task plans with verification criteria.
4. Execute Phase
/gsd:execute-phase <N>
Parallel agents write code. Orchestrator spawns specialists, collects results.
5. Verify Work
/gsd:verify-work [N]
Automated verification against goals and acceptance criteria.
6. Repeat → Complete → Next Milestone
/gsd:complete-milestone
/gsd:new-milestone [name]
Quick Mode
/gsd:quick [--full] [--discuss]
For smaller tasks — skip the ceremony.
Why It Works: 5 Engineering Pillars
1. Context Engineering
GSD manages Claude's context for you through structured documents:
| Document | Purpose |
|---|---|
PROJECT.md | Project definition and vision |
research/ | Research findings |
REQUIREMENTS.md | Detailed requirements |
ROADMAP.md | Milestone roadmap |
STATE.md | Current execution state |
PLAN.md | Active phase plan |
SUMMARY.md | Session summaries |
todos/ | Todo tracking |
Size limits based on where Claude's quality degrades. Stay under, get consistent excellence.
2. XML Prompt Formatting
Every plan is optimized XML for Claude:
<task type="auto">
<name>Create login endpoint</name>
<files>src/app/api/auth/login/route.ts</files>
<action>
Use jose for JWT. Validate credentials against users table.
Return httpOnly cookie on success.
</action>
<verify>curl -X POST localhost:3000/api/auth/login returns 200</verify>
<done>Valid credentials return cookie, invalid return 401</done>
</task>
Precise instructions. No guessing. Verification built in.
3. Multi-Agent Orchestration
The key to GSD's effectiveness: your main context window stays at 30-40%.
A thin orchestrator spawns specialized agents, collects results, and routes to the next step. Each phase runs in fresh subagent contexts — deep research, plan creation, code writing, automated verification — while your session stays fast and responsive.
4. Atomic Git Commits
Each task gets its own commit:
abc123f docs(08-02): complete user registration plan
def456g feat(08-02): add email confirmation flow
hij789k feat(08-02): implement password hashing
lmn012o feat(08-02): create registration endpoint
Every commit is surgical, traceable, and independently revertable. Git bisect can find exact failing tasks.
5. Modular by Design
- Add phases to current milestone
- Insert urgent work between phases
- Complete milestones and start fresh
- Adjust plans without rebuilding everything
Brownfield Support
Already have code? Map your existing codebase:
/gsd:map-codebase
Spawns parallel agents to analyze your stack, architecture, conventions, and concerns. Then /gsd:new-project knows your codebase — questions focus on what you're adding.
20+ Slash Commands
| Category | Commands |
|---|---|
| Core Workflow | new-project, discuss-phase, plan-phase, execute-phase, verify-work, audit-milestone, complete-milestone, new-milestone |
| Navigation | progress, help, update, join-discord |
| Brownfield | map-codebase |
| Phase Management | add-phase, insert-phase, remove-phase, list-phase-assumptions, plan-milestone-gaps |
| Session | pause-work, resume-work |
| Utilities | settings, set-profile, add-todo, check-todos, debug, quick, health |
Configuration
Core Settings
| Setting | Options | Default |
|---|---|---|
mode | yolo / interactive | interactive |
granularity | coarse / standard / fine | standard |
Model Profiles
Balance quality vs. token spend: quality, balanced, budget.
Workflow Agents
| Agent | Default | Purpose |
|---|---|---|
workflow.research | true | Deep research during planning |
workflow.plan_check | true | Plan verification |
workflow.verifier | true | Automated work verification |
workflow.auto_advance | false | Auto-advance to next phase |
Git Branching Strategies
| Strategy | Behavior |
|---|---|
none | Commits to current branch (default) |
phase | Branch per phase, merge at completion |
milestone | One branch per milestone |
GSD vs Alternatives
Category: This tool is a meta-prompting, context engineering & spec-driven development system for AI coding agents.
| Feature | GSD | BMAD Method | Aider |
|---|---|---|---|
| Focus | Solo builder workflow | Enterprise Agile AI | AI pair programming |
| Stars | 26K ⭐ | 39.6K ⭐ | 41.6K ⭐ |
| License | MIT | Other | Apache 2.0 |
| Language | JavaScript | JavaScript | Python |
| Target User | Solo devs & small teams | Enterprise teams | All developers |
| Context Engineering | ✅ Full system (PROJECT/REQUIREMENTS/ROADMAP/STATE/PLAN) | ✅ Constitution + Specs | ✅ Codebase mapping |
| XML Prompt Formatting | ✅ Optimized for Claude | ❌ | ❌ |
| Multi-Agent Orchestration | ✅ Subagent spawning | ✅ Specialized personas (PM, Architect, Dev) | ❌ |
| Spec-Driven Development | ✅ | ✅ Core philosophy | ❌ |
| Atomic Git Commits | ✅ Per-task | ❌ | ✅ Auto-commits |
| Main Context Stays Low | ✅ 30-40% | ❌ | ❌ |
| Brownfield Support | ✅ /gsd:map-codebase | ✅ | ✅ Codebase map |
| Quick Mode | ✅ | ❌ | ❌ |
| Claude Code Support | ✅ Primary | ✅ | ✅ |
| Multi-Runtime | ✅ Claude/OpenCode/Gemini/Codex | ✅ Multi-IDE | ✅ Multi-LLM |
| Model Profiles | ✅ Quality/Balanced/Budget | ❌ | ✅ Model selection |
| Git Branching Strategies | ✅ None/Phase/Milestone | ❌ | ❌ |
| Verification Phase | ✅ Built-in | ✅ | ❌ |
| Pause/Resume Sessions | ✅ | ❌ | ❌ |
| Install | ✅ npx one-liner | Manual setup | pip install |
| Enterprise Ceremonies | ❌ By design | ✅ Sprint/Stories/Retrospectives | ❌ |
| Team Size | Solo / Small | Medium / Large | Any |
When to choose GSD: You're a solo developer or small team who wants to describe ideas and have them built correctly — without enterprise theater. Context engineering keeps Claude focused, XML formatting eliminates guessing, multi-agent orchestration keeps your session fast, and atomic commits give surgical git history. Supports Claude Code, OpenCode, Gemini CLI, and Codex.
When to choose BMAD Method: You want a full enterprise Agile framework with specialized AI agent personas (PM, Architect, Developer, UX, Scrum Master). Sprint ceremonies, story points, stakeholder workflows. Better for larger teams. Note: higher workflow complexity.
When to choose Aider: You want AI pair programming in the terminal with git-native workflow, multi-LLM support (Claude, GPT, Gemini, Llama), and codebase mapping. No spec-driven workflow — more conversational, direct code editing. Apache 2.0 license.
Conclusion
Get Shit Done is the anti-enterprise answer to AI-assisted development. No sprint ceremonies. No story points. No Jira workflows. Just a brutally effective context engineering system that gives Claude Code everything it needs to build real software reliably. From XML-structured plans to multi-agent orchestration that keeps your context at 30-40%, from atomic git commits to brownfield codebase mapping — GSD is how solo developers and small teams ship consistently with AI coding agents.