spec-driven-development

10 repositories

Fission-AI/OpenSpec

Give AI coding assistants a spec to work from, not just a prompt — OpenSpec brings spec-driven development to AI coding assistants: you explore and propose changes via slash commands, and it generates proposals, requirements, design docs, and task checklists before code is written. Built for both new and existing codebases, it aims to scale from solo projects to enterprise teams.

TypeScriptAttention 40
★75,289
bmad-code-org/BMAD-METHOD

A sizing-aware method for agile AI-driven development — BMad Method is an agile framework for AI-driven development that scales itself to the task: small changes go straight to build, complex work gets fuller planning and architecture. It now ships two tracks—a stable v6.12 npm installer and a v7 alpha using a skills CLI—letting solo devs or small teams adopt either end-to-end or alongside existing workflows.

PythonAttention 38
★55,560
zeromicro/go-zero

A battle-tested Go framework now leaning into AI-native, spec-driven workflows — go-zero is a Go web/rpc framework with built-in resilience (circuit breaking, rate limiting, load shedding) proven at tens-of-millions-user scale. Its goctl tool generates code from simple API spec files across Go, iOS, Android, Kotlin, Dart, TypeScript and JavaScript, now expanding toward AI-native, spec-driven development.

GoAttention 31
★33,433
google/elephant-goldfish-model

Google's Plugin Splits Drafting and Review Roles for AI Coding Agents — EGM is an open-source plugin from Google that separates spec-writing from review: an "elephant" agent drafts with full project context, while "goldfish" reviewers examine the draft fresh, without prior conversation, to surface blind spots. It installs into Claude Code, Codex, and Antigravity, and includes bootstrap and drift reporting for existing codebases.

Python🇺🇸 North AmericaAttention 55
★21,484
open-gsd/gsd-core

A phase-loop framework that keeps AI coding agents' context clean — GSD Core drives AI coding agents like Claude Code, Codex, and Cursor through a disciplined Discuss-Plan-Execute-Verify-Ship loop, running heavy work in fresh-context subagents to fight context rot. Installable via npx, it suits developers who want spec-driven, multi-runtime AI coding workflows.

JavaScript🇺🇸 North AmericaAttention 49
★11,712
Gentleman-Programming/gentle-ai

A deterministic layer that gives your AI coding agent memory and proof — Gentle-AI configures 17 existing coding agents (Claude Code, Cursor, OpenCode, Codex, Pi, etc.) with persistent memory, Organic-Driven Development workflows, curated skills, MCP servers and personas. It adds evidence of what the agent did instead of asking you to trust it, and avoids locking you into one agent.

GoAttention 47
★9,416
JetBrains/thinkrail

JetBrains' new agentic IDE turns the pi coding agent into isolated worktree workspaces — ThinkRail wraps JetBrains' pi coding agent in an IDE that runs each task as an isolated git-worktree workspace, tracks project knowledge via living specs, and settles finished branches automatically. It's aimed at developers running parallel AI coding sessions who want spec-driven context instead of ad-hoc prompting, with extension support planned next.

TypeScript🇪🇺 EuropeAttention 34
★8,363
pullboard-dev/pullboard

Autonomous development for real products. You set the standard. Declare your vision. Walk away. It builds.

JavaScript🇺🇸 North America
★841
ericrisco/rsc-harness

A reviewable plan for setting up your coding agent's memory and skills — rsc-harness reads your project, asks about your goal and which assistants you use (18 targets including Claude Code, Codex, Cursor), then proposes a plan of skills, rules and MCP connections you approve before anything is written. It aims to avoid ad-hoc 'Frankenstein' agent setups by keeping project knowledge consistent across tools, installed via npx.

JavaScriptAttention 45
★640
mryo0826/aidd-practice-guide

Applying Scrum discipline to keep AI collaborators honest about guesses — A methodology guide for AI-Driven Development built around Claude Code and Agent Skills. It focuses not on tool usage but on process discipline: labeling claims by verification status, logging self-corrected errors, and tracking which chapters are confirmed versus unverified. Aimed at anyone involved in AI-assisted system development, including non-developers defining requirements.

★6