Most AI coding assistants excel at implementation but silently encode unstated decisions into code. This project treats AI as part of a delivery process: agents and named skills run structured workflows that make product, architecture, and testing decisions explicit and carry that context forward through planning and implementation.
What Sets It Apart
- Durable context and explicit decisions — workflows record the what/why behind changes so future agents or humans don’t re-ask the same questions; that reduces rework and drift in multi-agent pipelines.
- Right-sized planning tracks — three tracks (Quick Flow, BMad Method, Enterprise) let teams skip deep planning for trivial changes or add architecture, security, and compliance depth for larger initiatives, so the process matches the work.
- Skills-first integration — exposes commands (skills) inside AI coding tools and marketplaces (Claude Code, Codex, Skills CLI) so the method plugs into existing editor/tooling workflows and can be updated via npm/skills rather than a separate app.
- Two release lines with predictable stability tradeoffs — a frozen stable line (v6.12) for production stability and a v7 preview for rapid innovation and skill-driven capabilities.
Who it's for and tradeoffs
Great fit if you want to keep human judgment central while using AI to implement and verify changes, or if you need deliverables (PRDs, architecture briefs, tested builds) that carry forward. Also useful for teams adding AI-driven workflows to existing codebases. Look elsewhere if you only need a single-chat coding assistant or a standalone GUI IDE plugin without workflow governance. The method depends on Node.js, Python scripts, and an AI coding tool that supports skills; the v7 preview can introduce breaking changes and requires operational attention.
