What problem does it solve? AI coding agents often jump straight to implementation without a documented technical design, leading to inconsistent architecture decisions and untraceable rationale. This Skill enforces a design phase in a Spec-Driven Development (SDD) pipeline, producing a structured design.md that captures how a change will be implemented before any code is written. ## Core Features & Use Cases - Structured Design Documents: Produces a design.md with technical approach, architecture decisions (choice, alternatives, rationale), data flow diagrams, file change tables, interfaces, and testing strategy. - Multi-Mode Persistence: Supports engram, openspec, hybrid, and none artifact store modes, reading the proposal and spec artifacts and persisting the design accordingly. - Codebase-Grounded Design: Requires reading the actual affected code, existing patterns, and test infrastructure before designing, so decisions follow project conventions rather than generic best practices. - Use Case: An orchestrator delegates a change to this sub-agent after the proposal and spec phases; it reads the codebase, writes an architecture design under 800 words, persists it, and returns a summary ready for the task-breakdown phase. ## Quick Start Delegate the design phase for a change by invoking the sdd-design sub-agent with the change name and artifact store mode after the proposal and spec artifacts exist.
