open-gsd/gsd-core

Git. Ship. Done - Core

What it solves

It addresses "context rot"—the degradation of AI output quality that occurs as a coding session's context window fills up. It also solves the lack of shared memory between sessions and the absence of a verification step to ensure code actually works before it is shipped.

How it works

GSD Core implements a spec-driven development framework that guides AI coding agents through a disciplined five-step phase loop:

  1. Discuss: Capture implementation decisions.
  2. Plan: Research and decompose tasks to fit into fresh context windows.
  3. Execute: Run plans in parallel waves using subagents with clean 200k-token contexts.
  4. Verify: Diagnose and fix issues before completion.
  5. Ship: Create the PR and archive the phase.

To maintain state across sessions, it uses structured artifacts like STATE.md and CONTEXT.md.

Who it’s for

Developers using AI coding agents such as Claude Code, Cursor, Windsurf, Copilot, Codex, Kimi CLI, Antigravity CLI, and OpenCode.

Highlights

  • Cross-runtime compatibility: Works across a wide variety of AI coding tools.
  • Context Engineering: Prevents bloat by offloading heavy work to fresh-context subagents.
  • Structured Memory: Uses persistent artifacts to maintain project state across different sessions.
  • Disciplined Workflow: Enforces a research-plan-execute-verify-ship cycle.

Related

  • Project
    microsoft/agentrc

    A context engineering tool that automatically generates and maintains codebase-specific instructions and configurations to improve the performance of AI coding agents.

  • Project
    jsmastery-pro/skills

    A set of modular AI agent skills that implement a professional engineering workflow, moving project state from chat sessions into repository files for persistence and team collaboration.

  • Dispatch
    The Pitfalls of AI Coding Assistants: Managing Context, Bloat, and Architectural Integrity

    Developers are reporting significant challenges with AI coding assistants, including codebase bloat, context window limitations, and a lack ofabsence of holistic system awareness, requiring a shift from 'hands-off' delegation to strict human-led architectural oversight.

  • Dispatch
    Optimizing AI Coding Agents with Spec-Driven Development

    Explore a new Spec-Driven Development workflow for Claude Code that emphasizes decomposition, context management, and persistent specs to improve agent performance and reduce costs.

  • Project
    NeoLabHQ/context-engineering-kit

    A plug‑in marketplace for LLM‑based coding assistants (Claude Code, Gemini, Antigravity, etc.) that supplies token‑efficient, context‑engineering patterns—reflection loops, sub‑agent judges, specification‑first development, code‑review agents, and Git helpers—to make AI‑generated code more reliable and easier to integrate into CI pipelines.