SDD-Driven Development Workflow

Problem

AI-assisted requirement development often fails at traceability: requirements are interpreted ad hoc, technical plans drift from the codebase, and changes lack acceptance evidence. SDD-Driven Development Workflow targets projects that have a PRD, optional SDD documents, and a code repository, and structures AI development into reviewable stages instead of jumping directly from request to code edits.

How It Works

The skill anchors each requirement in a REQ_ROOT directory and archives inputs such as PRDs under input_materials/. Key steps include:
- Analysis: produce a change-point list and distinguish implemented, pending, and unchanged behavior, then confirm completeness with a human.
- Design: map the technical plan to each change point, and load company or generic framework notes when components are involved.
- Review: iterate the plan, generate external documents, and archive the SDD gap list.
- Implementation and verification: execute tasks, close T* and AC-* items, and optionally run AI-assisted CR or test-case validation.

Fit and Cautions

This workflow is human-gated at critical decisions. It can work from PRD and code exploration without SDD, but interface contracts, RPC field compatibility, and high-risk changes must be explicitly documented. It fits teams that already maintain requirement docs and code, and want auditable evidence from AI-assisted development.