SDD Builder AI — Govern AI Coding with Spec-Driven Development

v1.0.0 | Stop losing control of AI-generated code. Ship with confidence.

SDD Builder AI is a VS Code extension that brings Spec-Driven Development to your AI coding workflow. It sits between your ideas and your AI agent: creates specification documentation and breaks them into small, scoped, reviewable specs that keep every execution focused, traceable, and under your control.

The Problem

Every developer using AI coding tools has hit this wall:

• Ask for one feature → the agent rewrites 15 files
• Large codebase → context overflow → the agent forgets critical code halfway through
• Agent invents APIs, imports, and patterns that don't exist
• Massive diffs you can't reasonably review
• No record of what was asked, what was built, or why

You're not bad at prompting. The tooling is missing a governance layer.

The Solution

SDD Builder AI introduces a structured workflow between your ideas and the AI. The objective is to break down development into small, focused specs that the AI can execute with precision and that you can review with confidence.

Requirements/Ideas → AI engineers Requirements → AI Creates Spec Cards → Execute with AI → Review Diff → Ship

Every AI execution is governed by a spec file — a small contract that tells the agent exactly:

• What to build (and nothing else)
• Which files to read for context (no guessing, no context overflow)
• Which files are off-limits (no accidental rewrites)
• What "done" looks like (acceptance criteria the AI must verify)

The result: small, focused diffs you can review in minutes, not hours. Reuse and standardize your AI prompt for development

Bring Your Own AI (BYOK)

SDD Builder AI does not manage API keys or subscriptions. It orchestrates the AI CLI tools you already use:

• Claude CLI — Anthropic's command-line agent (claude)
• GitHub Copilot CLI — GitHub's AI assistant (gh copilot)

Install and authenticate with your own account. Your keys, your billing, your model choice — the extension assembles context and delegates execution. It never touches your credentials.

Key Features

Project Initialization

Run SDD: Initialize Project to scaffold the standard .sdd/ folder structure:

.sdd/
  product_requirements/   # Feature requirements and idealizations
  execution/              # Execution logs and audit records
  skills/                 # Agent skill files (.md) per technology tag
  ai-config.json          # AI provider and execution configuration
  config.json             # Project settings
  conventions.md          # Your coding standards (optional but recommended)
.specs/                   # All .sdd.md spec files

Overview dashboard

High-level view of your project's development health at a glance.

• Spec counts by status (Draft, Ready, In Progress, Review, Done)
• Completion percentage across all specs
• SDD tags overview and recent activity feed
• All execution logs saved in .sdd/execution/ for full audit after development

Requirement Board

Capture raw ideas and turn them into structured development plans with AI.

• Add feature ideas with a name, description, and acceptance criteria
• AI Idealization — the AI acts as a software architect, producing a concise requirements document from your idea
• AI Spec Generation — one click decomposes the idealization into dependency-ordered SDD spec cards ready for development

SDD Kanban Board

A five-column board where each spec moves from idea to shipped code.

Drag and drop cards between columns. Every transition is validated — no skipping stages.

Bulk Execution

Select multiple Ready specs and execute them sequentially in one action.

• Real-time progress tracking per card — spinner, checkmark, or failure indicator
• Stops automatically on failure so you can review before continuing
• Only available in full-permission AI modes to prevent mid-execution interruptions

Request Changes Flow

When a spec needs rework, your feedback is injected directly into the next execution.

• Review diff in split-view (spec on left, changes on right)
• Approve to mark Done, or write feedback to request changes
• Feedback is appended to the spec's context section and status resets to Ready
• The AI receives your exact feedback in the next run — no context is lost

Scoped Execution — AI That Respects Your Architecture

Each spec enforces hard boundaries:

• 3 files max to create or edit (excluding tests) — no surprise rewrites
• relevant_files gives the agent exactly the context it needs — zero context overflow
• must_not_touch protects sensitive files — auth, config, database schemas stay safe
• conventions.md injects your project's coding standards into every execution
• Skills per spec tag — tag specs with python, react, database, etc., and configure skill files that are injected into the execution prompt automatically

AI Configuration

Separate configuration for requirements AI and execution AI.

• Choose your AI provider: Claude CLI or GitHub Copilot CLI
• Set permission modes per provider
  • Full permission: a terminal is created and the AI CLI executes all commands. The watch sentinel monitors when the AI finishes and update the SDD card for review.
  • Plan mode: a terminal opens and the AI CLI executes in plan mode. The user should press the complete button to move it to review phase.
• Tag-to-skill mappings with custom skill files. The skills are injected into the prompt when a spec with the corresponding tag is executed, giving you precise control over the agent's capabilities per task.
• Prompt templates for idealization, spec generation, and execution phases — customize the instructions and context you give the AI at each step.
• Configurable git commit and git pull-request commands into the prompt context.

Why Specs Beat Prompts

Requirements

• VS Code 1.85 or later
• AI CLI (BYOK) — one of the following, installed and authenticated:
  • Claude CLI (npm install -g @anthropic-ai/claude-code)
  • GitHub Copilot CLI (gh extension install github/gh-copilot)
• Git configured in the workspace
• gh CLI (optional, for PR automation) — cli.github.com

Getting Started

1. Install the extension from the VS Code Marketplace
2. Open a project folder in VS Code
3. Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) and run SDD: Initialize Project
4. Click the SDD icon in the Activity Bar to open the sidebar
5. Click Open Dashboard to see the SDD Builder AI dashboard
6. Go to the Requirement Board and add your first feature idea
7. Click Idealize to let the AI structure your requirement
8. Click Create SDD Cards to generate spec files
9. Open the Kanban Board, move a spec to Ready, and execute

Tech Stack

Feedback & Issues

Report bugs and feature requests at GitHub Issues.

SDD Builder AI — Structure your AI. Ship with confidence.