Part VII: Spec Kit Workflow
The Three Commands That Transform SDD into Daily Practice
Spec-Driven Development is a methodology—a philosophy of putting specifications first. But methodology alone doesn't ship features. You need workflow: concrete commands you run, artifacts you produce, and a pipeline that takes you from "I want to build X" to "X is built and tested."
Part VII introduces the Spec Kit workflow—the three commands that transform SDD from methodology to daily practice:
/speckit.specifycreates feature specifications from simple descriptions/speckit.plangenerates implementation plans from specifications/speckit.tasksproduces executable task lists from plans
Together, these commands form the core SDD pipeline: a repeatable, AI-assisted workflow that takes you from a one-sentence feature idea to an ordered list of atomic, testable tasks ready for implementation.
What You Will Learn
Chapter 20: /speckit.specify — Feature Specification
You will learn how /speckit.specify transforms a simple feature description into a complete specification. The command automates feature numbering (scanning existing specs for the next number), branch creation (generating semantic branch names), and template-based generation (copying and customizing the feature spec template). All documents live in specs/[branch-name]/. You will understand the specification template structure—how it constrains AI to focus on WHAT users need and WHY, avoid HOW to implement, mark ambiguities with [NEEDS CLARIFICATION], and include completeness checklists. A hands-on tutorial walks you through creating a "real-time chat" feature specification from start to finish.
Chapter 21: /speckit.plan — Implementation Planning
You will learn how /speckit.plan transforms a specification into architecture and implementation plan. The command reads spec.md and produces plan.md, data-model.md, contracts/, research.md, and quickstart.md. You will understand specification analysis (requirements, user stories, acceptance criteria), constitutional compliance (alignment with project constitution), and technical translation (converting business requirements into technical decisions). The plan template includes technology decisions with rationale, data models with field definitions, API contracts with request/response schemas, phase gates (Simplicity, Anti-Abstraction, Integration-First), and file creation order (contracts → tests → source). A tutorial guides you through generating an implementation plan for the "real-time chat" feature.
Chapter 22: /speckit.tasks — Task Generation
You will learn how /speckit.tasks analyzes plan and design documents to generate an executable task list. The command reads plan.md (required), data-model.md, contracts/, and optionally research.md, producing tasks.md with atomic, ordered, traceable tasks. You will understand task derivation (converting contracts, entities, scenarios into specific tasks), parallelization (marking independent tasks [P] and outlining safe parallel groups), and task anatomy (what makes a task atomic, testable, traceable). A tutorial walks you through generating tasks for the "real-time chat" feature, identifying parallelizable work, and executing the first task with AI. You will see the complete pipeline in action and compare traditional vs. SDD workflow time (12 hours vs. 15 minutes).
The Connection
The three chapters form a linear pipeline:
- Chapter 19 establishes the specification—the WHAT and WHY that anchors everything that follows.
- Chapter 20 translates specification into architecture—the HOW, with data models, contracts, and phase gates.
- Chapter 21 breaks the plan into executable units—the TASKS that you or an AI agent will implement one by one.
Each command consumes the output of the previous. Specification feeds plan. Plan feeds tasks. Tasks feed implementation. By the end of Part VII, you will have mastered the core SDD pipeline and be able to take any feature idea from concept to executable task list in minutes rather than hours.