GSD Core
A context-engineering and spec-driven development system that runs AI coding agents through a five-step discuss, plan, execute, verify and ship loop to keep long sessions from losing quality.
GSD Core is a lightweight framework that drives AI coding agents, including Claude Code, OpenCode, Codex and others, through a disciplined phase loop rather than one long conversation. It targets context rot, the quality decline that builds up as an agent's context window fills, by running research, planning and execution work in fresh-context subagents while keeping the main session lean.
Each project milestone repeats a five-step loop: discussing implementation decisions, planning and decomposing the work so each piece fits a fresh context window, executing plans in parallel with a clean context per executor, verifying what was built before declaring it done, and shipping a pull request before moving to the next phase.
It installs through a Node-based installer that detects the current agent runtime and either the current project or a global location.