
Get Shit Done
github.com/gsd-build/get-shit-done- Category
- Developer Tools
- Rank
- No. 2016Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- gsd-build
- GitHub
- 64.6k stars
- Latest release
- v1.43.0-rc2
- Date
About
A lightweight meta-prompting and context engineering system that prevents context rot in Claude Code and other AI coding assistants. Uses spec-driven development with built-in quality gates to maintain code quality as the AI's context window fills up.
What it does
It turns a software idea into a managed sequence of questions, research, requirements, phase plans, implementation, and verification. Specialized agents handle planning, execution, checking, security, documentation, and other focused jobs. The workflow stores decisions and progress as project artifacts, groups independent work into parallel waves, and creates atomic commits for completed tasks.
Why it's ranked here
The workflow is unusually explicit about traceability and verification. Requirements map to roadmap phases, plans receive pre-execution checks, and completed work is tested against phase goals rather than task completion alone. However, this repository is archived and redirects current development elsewhere, so it is useful for understanding the system but is the wrong source for current releases, issues, or contributions.
What's good
Its quality gates address concrete failure modes. The planner breaks work into context-sized tasks, checks dependencies and reachable references, and can require human approval before suspicious package installs. Executors report deviations and commit tasks separately. Verification audits requirement coverage, cross-phase integration, and whether tests genuinely prove their claims, including skipped tests, weak assertions, and circular expected values.
Tradeoffs
The full workflow carries substantial ceremony: research artifacts, requirements, roadmaps, phase discussions, plans, summaries, verification reports, and state tracking. That structure suits consequential projects better than tiny edits, for which separate fast and quick modes exist. Some capabilities depend on specific runtimes, while beta cloud planning is Claude Code only and may change or disappear. Preview packages may regress and are explicitly discouraged for production delivery.
How to use it well
Use it for multi-step engineering work where requirements, decisions, implementation, and proof must remain connected across long sessions. Start with project discovery, discuss each phase, then plan, execute, verify, and ship. Choose fast or quick modes for narrow tasks that do not justify the full artifact chain. Treat it as an orchestration and quality framework, not a replacement for your coding runtime, package registry checks, or human acceptance testing.
Technical notes+
README.md marks this checkout as archived and redirects active work to GSD Core. package.json defines a Node.js package requiring Node 22 or newer, exposes get-shit-done-cc, gsd-sdk, and gsd-tools binaries, and depends on the Claude Agent SDK plus WebSocket support. docs/CLI-TOOLS.md describes the legacy CommonJS CLI, typed SDK registry dispatch, strict SDK mode, explicit subprocess fallback, and structured dispatch events. docs/json-errors.md specifies single-line JSON failures on stderr with stable reason codes. docs/COMMANDS.md documents runtime-specific command spellings, while docs/AGENTS.md and docs/INVENTORY.md describe the specialized multi-agent surface.
Observed
- License
- MIT
- Package surface
- Node.js npm package with three command-line binaries
- Runtime requirement
- Node.js 22 or newer
- Languages
- JavaScript and TypeScript surfaces
- Interfaces
- Runtime commands, command-line tools, and a typed SDK
- Runtime support
- Documentation covers Claude Code, Gemini CLI, Codex, Copilot, OpenCode, Kilo, and additional coding runtimes
- Repository status
- Archived redirect; active development continues in the Open GSD repository
- Testing structure
- Package scripts define unit, integration, install, security, slow, and coverage suites
Read from README.md, package.json, docs/BETA.md, docs/AGENTS.md, docs/CANARY.md, docs/README.md, docs/COMMANDS.md, docs/FEATURES.md, docs/CLI-TOOLS.md, docs/INVENTORY.md, docs/USER-GUIDE.md, docs/json-errors.md.
What it can do
Prevent context rot in AI coding assistants
AI coding session with growing context → Maintained context quality and relevance
Format prompts using XML structure
Natural language prompts → Structured XML-formatted prompts
Engineer context for AI coding sessions
Code requirements and existing context → Optimized context for AI assistant
Implement spec-driven development workflows
Project specifications and requirements → Structured development workflow
Apply quality gates to maintain code standards
Generated code from AI assistant → Quality-checked and validated code
Manage AI assistant context windows
Ongoing AI coding conversation → Optimized context window utilization
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.