AI‑assisted Coding Workflows

• Classic human‑led docs (PRD → design/tech plan → tasks + ADRs) Keep PRDs, design/tech plans, and a task backlog in the repo; capture key choices as short ADRs that AIs can reference.

• Spec-Driven Development (GitHub Spec Kit — https://github.com/github/spec-kit) Write a plain‑English spec, generate a technical plan, break it into tasks, and implement step‑by‑step with an AI assistant (spec → plan → tasks → implement). Also see: The ONLY guide you’ll need for GitHub Spec Kit. Tools that support this workflow: Kiro.

• Taskmaster (PRD → tasks → guided execution — https://www.task-master.dev/) Start from a human‑written PRD; a CLI parses it into an ordered, dependency‑aware task list that an AI follows one task at a time.

• BMAD Method (multi‑agent “agile team” — https://github.com/bmad-dev/bmad) Assign roles (analyst, PM, architect, scrum master, dev, QA). Planning agents produce PRD + architecture; development agents turn “story files” into code under human review.

• MetaGPT (software‑company‑in‑a‑box — https://github.com/geekan/MetaGPT) Given one‑line requirements, a multi‑agent pipeline outputs PRD, design docs, APIs, tasks, and initial code/repo.

• ChatDev (communicative multi‑agent SDLC — https://github.com/OpenBMB/ChatDev) Role‑based agents collaborate through staged chats across design → coding → testing, producing docs and code incrementally.