cc-sdd

Why we included this project

Rather than letting an AI coding agent improvise each request from a blank page, this project installs a spec-driven workflow into whichever agent your team already uses. It runs discovery, requirements, design, and task decomposition before handing work to a long-running implementer that works in TDD cycles behind a feature flag, with each task reviewed independently and an auto-debug pass that digs into root causes when the implementer gets blocked or the reviewer keeps rejecting. A lead can approve the spec once and let implementation run on its own, then pick the workstream back up cleanly after interruptions. Because the same 17-skill setup installs across Claude Code, Codex, Cursor, Copilot, and Windsurf, teams get consistent delivery without building their own agent orchestration.