Get Shit Done: Meta-Prompting System for AI Coding Agents
Get Shit Done (GSD) is a meta-prompting, context engineering, and spec-driven development system designed to work with multiple AI coding agents. The tool focuses on solving context rot—the quality degradation that occurs when AI agents fill their context windows with irrelevant information.
Supported Platforms and Installation
The system works with Claude Code, OpenCode, Gemini CLI, Codex, Copilot, and Antigravity. Installation is straightforward:
npx get-shit-done-cc@latestThe installer prompts you to choose runtime and installation location (global or local). For non-interactive installations in Docker, CI, or scripts:
- Claude Code:
npx get-shit-done-cc --claude --global(installs to ~/.claude/) - OpenCode:
npx get-shit-done-cc --opencode --global(installs to ~/.config/opencode/) - Gemini CLI:
npx get-shit-done-cc --gemini --global(installs to ~/.gemini/) - Codex:
npx get-shit-done-cc --codex --global(installs to ~/.codex/) - Copilot:
npx get-shit-done-cc --copilot --global(installs to ~/.github/) - Antigravity:
npx get-shit-done-cc --antigravity --global(installs to ~/.gemini/antigravity/) - All runtimes:
npx get-shit-done-cc --all --global
How It Works
GSD uses context engineering, XML prompt formatting, subagent orchestration, and state management behind the scenes. The creator describes it as having "complexity in the system, not in your workflow." Users see only a few commands that work consistently.
The system extracts what it needs to know from user descriptions, then lets the AI agent do the work with built-in verification. Codex installation uses skills (skills/gsd-*/SKILL.md) rather than custom prompts.
Verification Commands
After installation, verify with these commands depending on your runtime:
- Claude Code / Gemini:
/gsd:help - OpenCode:
/gsd-help - Codex:
$gsd-help - Copilot:
/gsd:help - Antigravity:
/gsd:help
The creator built GSD as a solo developer who doesn't write code directly—Claude Code does. They found other spec-driven development tools too complex with "sprint ceremonies, story points, stakeholder syncs, retrospectives, Jira workflows" or lacking big-picture understanding.
GSD is positioned as an alternative to what the creator calls "enterprise theater"—tools that simulate 50-person engineering organizations when users just need to build things that work.
📖 Read the full source: HN AI Agents
👀 See Also
Klaw.sh: Kubernetes-Style Orchestration for AI Agents
Klaw.sh provides an orchestration solution for AI agent deployment, modeled on Kubernetes. It simplifies management with clusters, namespaces, and channels, achieving memory reduction by rewriting from Node.js to Go.
GSD-Lite: A State Machine for Claude Code That Enforces TDD and Prevents Test Skipping
GSD-Lite is an open-source MCP server that adds a 12-state workflow machine to Claude Code, enforcing test-driven development with specific anti-rationalization prompts and separate agent contexts for execution, review, and debugging.
RTX 5060 Ti 16GB Local LLM Benchmarks: 30B Models Still Lead for Coding
Benchmarks on an RTX 5060 Ti 16GB show Unsloth Qwen3-Coder-30B UD-Q3_K_XL achieving 76.3 tok/s on Ubuntu with quality score 8.14, making it the recommended default coding model. The Unsloth Qwen3.5-35B UD-Q2_K_XL hits 80.1 tok/s but with lower quality scores.
Krasis: Hybrid CPU/GPU Runtime for Large MoE Models Achieves 3,324 tok/s Prefill on RTX 5080
Krasis is a hybrid CPU/GPU runtime that runs large MoE models by handling prefill on GPU and decode on CPU, achieving 3,324 tokens/second prefill on an RTX 5080 with Qwen3-Coder-Next 80B Q4. It requires ~2.5x model size in system RAM but enables running models too large for VRAM.